id,node_id,number,title,user,state,locked,assignee,milestone,comments,created_at,updated_at,closed_at,author_association,active_lock_reason,draft,pull_request,body,reactions,performed_via_github_app,state_reason,repo,type
614275938,MDU6SXNzdWU2MTQyNzU5Mzg=,4045,Millisecond precision is lost on datetime64 during IO roundtrip,10676434,closed,0,,,9,2020-05-07T19:01:18Z,2021-01-03T23:39:04Z,2021-01-03T23:39:04Z,NONE,,,,"<!-- A short summary of the issue, if appropriate -->
I have millisecond-resolution time data as a coordinate on a DataArray. That data loses precision when round-tripping through disk.

#### MCVE Code Sample
<!-- In order for the maintainers to efficiently understand and prioritize issues, we ask you post a ""Minimal, Complete and Verifiable Example"" (MCVE): http://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports -->

[bug_data.p.zip](https://github.com/pydata/xarray/files/4595145/bug_data.p.zip)

Unzip the data. It will result in a pickle file.

```python
bug_data_path = '/path/to/unzipped/bug_data.p'
tmp_path = '~/Desktop/test.nc'

with open(bug_data_path, 'rb') as f:
    data = pickle.load(f)

selector = dict(animal=0, timepoint=0, wavelength='410', pair=0)

before_disk_ts = data.time.sel(**selector).values[()]

data.time.encoding = {'units': 'microseconds since 1900-01-01', 'calendar': 'proleptic_gregorian'}

data.to_netcdf(tmp_path)
after_disk_ts = xr.load_dataarray(tmp_path).time.sel(**selector).values[()]

print(f'before roundtrip: {before_disk_ts}')
print(f' after roundtrip: {after_disk_ts}')
```
output:
```
before roundtrip: 2017-02-22T16:24:10.586000000
after roundtrip:  2017-02-22T16:24:10.585999872
```

#### Expected Output
```
Before: 2017-02-22T16:24:10.586000000
After:  2017-02-22T16:24:10.586000000
```

#### Problem Description
<!-- this should explain why the current behavior is a problem and why the expected output is a better solution -->

As you can see, I lose millisecond precision in this data. (The same happens when I use millisecond in the encoding).

#### Versions

<details><summary>Output of <tt>xr.show_versions()</tt></summary>

<!-- Paste the output here xr.show_versions() here -->
INSTALLED VERSIONS
------------------
commit: None
python: 3.7.6 | packaged by conda-forge | (default, Jan  7 2020, 22:05:27) 
[Clang 9.0.1 ]
python-bits: 64
OS: Darwin
OS-release: 19.4.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: None.UTF-8
libhdf5: 1.10.5
libnetcdf: 4.7.3

xarray: 0.15.1
pandas: 1.0.1
numpy: 1.18.1
scipy: 1.4.1
netCDF4: 1.5.3
pydap: None
h5netcdf: 0.8.0
h5py: 2.10.0
Nio: None
zarr: None
cftime: 1.0.4.2
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2.11.0
distributed: 2.14.0
matplotlib: 3.1.3
cartopy: None
seaborn: 0.10.0
numbagg: None
setuptools: 45.2.0.post20200209
pip: 20.0.2
conda: None
pytest: 5.3.5
IPython: 7.12.0
sphinx: 2.4.3

</details>
","{""url"": ""https://api.github.com/repos/pydata/xarray/issues/4045/reactions"", ""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,completed,13221727,issue
613717463,MDU6SXNzdWU2MTM3MTc0NjM=,4042,DataArray coordinates transformed into variables when saved to disk,10676434,closed,0,,,1,2020-05-07T01:51:36Z,2020-05-07T18:42:52Z,2020-05-07T18:42:52Z,NONE,,,,"<!-- A short summary of the issue, if appropriate -->
When I save my DataArray to disk using ``to_netcdf``, then try to reload it as a DataArray, it fails the roundtrip. When I load it as a DataSet, I see that the coordinates have been transformed into variables.

Also, the attributes have disappeared.

#### MCVE Code Sample
<!-- In order for the maintainers to efficiently understand and prioritize issues, we ask you post a ""Minimal, Complete and Verifiable Example"" (MCVE): http://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports -->
[bug_data.p.zip](https://github.com/pydata/xarray/files/4590243/bug_data.p.zip)

Unzip the file. It should be a pickle file.

```python
import pickle
import xarray as xr

data = pickle.load(open('/path/to/bug_data.p', 'rb'))

print(type(data) == xr.DataArray)
# >>> True

path = '/var/tmp/bug_data.nc'

data.to_netcdf(path, format='NETCDF4', mode='w')

# xr.open_dataarray(path) # fails due to multiple variables

xr.open_dataset(path) # succeeds
```
<img width=""717"" alt=""image"" src=""https://user-images.githubusercontent.com/10676434/81245480-add0b480-8fda-11ea-9e74-1e0bc420b41d.png"">


#### Expected Output
I expect ``xr.open_dataarray(path)`` to succeed, and for the result to be equal to ``data``.

#### Problem Description
<!-- this should explain why the current behavior is a problem and why the expected output is a better solution -->

The DataArray -> Disk -> DataArray roundtrip should be seamless.

#### Versions

<details><summary>Output of <tt>xr.show_versions()</tt></summary>

<!-- Paste the output here xr.show_versions() here -->
INSTALLED VERSIONS
------------------
commit: None
python: 3.7.6 | packaged by conda-forge | (default, Jan  7 2020, 22:05:27) 
[Clang 9.0.1 ]
python-bits: 64
OS: Darwin
OS-release: 19.4.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: en_US.ISO8859-1
libhdf5: 1.10.5
libnetcdf: 4.7.3

xarray: 0.15.1
pandas: 1.0.1
numpy: 1.18.1
scipy: 1.4.1
netCDF4: 1.5.3
pydap: None
h5netcdf: 0.8.0
h5py: 2.10.0
Nio: None
zarr: None
cftime: 1.0.4.2
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2.11.0
distributed: 2.14.0
matplotlib: 3.1.3
cartopy: None
seaborn: 0.10.0
numbagg: None
setuptools: 45.2.0.post20200209
pip: 20.0.2
conda: None
pytest: 5.3.5
IPython: 7.12.0
sphinx: 2.4.3

</details>
","{""url"": ""https://api.github.com/repos/pydata/xarray/issues/4042/reactions"", ""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,completed,13221727,issue
605717342,MDU6SXNzdWU2MDU3MTczNDI=,3997,DataArray.to_netcdf breaks in filename-dependent manner,10676434,closed,0,,,3,2020-04-23T17:31:08Z,2020-05-07T02:10:44Z,2020-05-07T02:10:44Z,NONE,,,,"<!-- A short summary of the issue, if appropriate -->
Writing a DataArray to disk breaks if the filename is `test.nc`. I'm unsure if other names pose a similar problem.

#### MCVE Code Sample
<!-- In order for the maintainers to efficiently understand and prioritize issues, we ask you post a ""Minimal, Complete and Verifiable Example"" (MCVE): http://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports -->

```python
# load data into `imgs`
>>> type(imgs) == xr.core.dataarray.DataArray
True

>>> filename = '/Users/sean/Desktop/test.nc'
>>> imgs.to_netcdf(filename)
>>> xr.open_dataarray(filename)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-132-9eaec208384d> in <module>
      2 filename = '/Users/sean/Desktop/test.nc'
      3 imgs.to_netcdf(filename)
----> 4 xr.open_dataarray(filename)

~/opt/anaconda3/envs/pharedox/lib/python3.7/site-packages/xarray/backends/api.py in open_dataarray(filename_or_obj, group, decode_cf, mask_and_scale, decode_times, autoclose, concat_characters, decode_coords, engine, chunks, lock, cache, drop_variables, backend_kwargs, use_cftime)
    676     if len(dataset.data_vars) != 1:
    677         raise ValueError(
--> 678             ""Given file dataset contains more than one data ""
    679             ""variable. Please read with xarray.open_dataset and ""
    680             ""then select the variable you want.""

ValueError: Given file dataset contains more than one data variable. Please read with xarray.open_dataset and then select the variable you want.

>>> filename = '/Users/sean/Desktop/testing.nc'
>>> imgs.to_netcdf(filename)
>>> xr.open_dataarray(filename)
```
![image](https://user-images.githubusercontent.com/10676434/80130304-f9f11300-855d-11ea-9ba3-1730068a6dca.png)

#### Expected Output
![image](https://user-images.githubusercontent.com/10676434/80129938-70414580-855d-11ea-9f93-a83a878b0be2.png)


#### Problem Description
<!-- this should explain why the current behavior is a problem and why the expected output is a better solution -->

Saving the DataArray to a file called `test.nc` seems like it generates a DataSet with multiple variables? Or otherwise destroys the expected formatting for `xr.load_datarray`.


#### Versions

<details><summary>Output of <tt>xr.show_versions()</tt></summary>

<!-- Paste the output here xr.show_versions() here -->
INSTALLED VERSIONS
------------------
commit: None
python: 3.7.6 | packaged by conda-forge | (default, Jan  7 2020, 22:05:27) 
[Clang 9.0.1 ]
python-bits: 64
OS: Darwin
OS-release: 19.4.0
machine: x86_64
processor: i386
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: None.UTF-8
libhdf5: 1.10.5
libnetcdf: 4.7.3

xarray: 0.15.1
pandas: 1.0.1
numpy: 1.18.1
scipy: 1.4.1
netCDF4: 1.5.3
pydap: None
h5netcdf: 0.8.0
h5py: 2.10.0
Nio: None
zarr: None
cftime: 1.0.4.2
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: None
dask: 2.11.0
distributed: 2.14.0
matplotlib: 3.1.3
cartopy: None
seaborn: 0.10.0
numbagg: None
setuptools: 45.2.0.post20200209
pip: 20.0.2
conda: None
pytest: 5.3.5
IPython: 7.12.0
sphinx: 2.4.3

</details>
","{""url"": ""https://api.github.com/repos/pydata/xarray/issues/3997/reactions"", ""total_count"": 0, ""+1"": 0, ""-1"": 0, ""laugh"": 0, ""hooray"": 0, ""confused"": 0, ""heart"": 0, ""rocket"": 0, ""eyes"": 0}",,completed,13221727,issue