Deprecate Dataset.T as an alias for Dataset.transpose()

[shoyer]

• Closes Remove the Dataset.T property? #1232
• Tests added / passed
• Passes git diff upstream/master | flake8 --diff
• Fully documented, including whats-new.rst for all changes and api.rst for new API

[rabernat]

This is great. I hit this annoyance all the time with any dataset from the IRI data library:
http://iridl.ldeo.columbia.edu/

The time dimension is always called T.

[brian-rose]

I'm happy to see this too!
I raised this way back in #687, but at the time the consensus was that .T notation was preferred for transposing because of consistency with numpy.
I've been using .rename on all variables called T ever since -- usually temperature in my use cases.

[max-sixty]

Good to go back and read my own comments:

Removing .T would break so much existing code, I don't think that's going to get much support. It's also consistent with numpy.

Good case to reflect on whether we're looking at the idea rather than the person's experience; whether we're using the right cognitive shortcuts ('[not] going to get much support'); etc