Raise on inplace=True

[max-sixty]

• Removes inplace, and relatedly closes can't do in-place clip() with DataArrays. #1997 (and maybe Document that Dataset.update defaults to being inplace. #2951)
• Passes black . && mypy . && flake8
• Fully documented, including whats-new.rst for all changes and api.rst for new API
• Remove from doc strings, docs

Do we want to raise an explicit error for another minor release (rather than remove all traces of inplace)? I'd vote +0.5 - it's likely that some people don't check deprecation warnings and this states the problem clearly

What do we want to do re .update (#2951)?

[shoyer]

It would also be nice to also update these methods to avoid using the inplace argument internally (e.g., to replace it with a literal Boolean if necessary).

[shoyer]

This should probably be TypeError, which is what Python uses for unrecognized arguments.

[max-sixty]

Great - what are your thoughts on .inplace @shoyer ?

[max-sixty]

All done, unless anyone has thoughts re inplace

[shoyer]

Could we make _check_inplace() entirely non-assigning? e.g., inplace = _check_inplace(inplace) -> _check_inplace(inplace)?

[max-sixty]

Could we make _check_inplace() entirely non-assigning? e.g., inplace = _check_inplace(inplace) -> _check_inplace(inplace)?

Yes, good idea.

Currently ds.update(inplace=[True|False]) will raise, by default it'll be True, is that OK?

[shoyer]

Yes, ds.update() should always be inplace.

[shoyer]

thanks @max-sixty !