Pandas 1.2.0 compatibility - `std()` for the `datetime64[ns]` dtype

This doesn’t work yet in dask because here the implementation of std generally depends on var, which doesn’t work for datetimes.

Some tests are skipped for this in dask/dataframe/tests/test_arithmetics_reduction.py

[bglossner]

Sorry for the long break! Created a PR when you have the time. It addresses the dtype error I was getting and also has code for doing axis=1.

[bglossner]

Hi there again! Sorry for not getting back to this sooner.

This isn’t done but I’d rather not do more work on it until I have some feedback on how I am doing it, since I think y’all may have some good feedback to where I should be doing things differently.

Here’s the commit: https://github.com/bglossner/dask/commit/922d7a1e28a7452153d3b72f54aa48f91d871acc. I’ll probably have more questions after this. Thanks a bunch and let me know if you’d rather have a PR instead!

I originally included timedelta in this as well but figured that might be feature creeping. Hopefully not hard to add, though (was just adding np.timedelta64 to the include for what I already had).