End of deprecation cycle: default to using new combine kwargs

[jsignell]

It's been almost 6 months and I haven't seen any reports of issues. All the original PR (#10062) did was raise FutureWarnings though, so this could still have an impact on people who ignored those.

• Closes Stricter defaults for concat, combine, open_mfdataset, merge #8778
• Tests added
• User visible changes (including notable bug fixes) are documented in whats-new.rst
• New functions/methods are listed in api.rst

[jsignell]

I am debating whether to make use_new_combine_kwarg_defaults a no-op. On the one hand it is a useful testing loop to run your combine logic with the old defaults and the new defaults and figure out if there are any differences. On the other hand what if people just set it to False and never change it?

By making it a no-op I don't mean removing the option, just enforcing that if it's set it is True. This would allow us to remove some code and warnings as well as the tests that exercise the option.

[jsignell]

I think we should put the specific version in this message, but this is just a guess at what that version will be.

[dcherian]

This change had a big blast radius (every user using open_mfdataset!); so I propose we wait another 6 months.

[jsignell]

This change had a big blast radius (every user using open_mfdataset!); so I propose we wait another 6 months.

Yeaaahh fair enough.

[jsignell]

Actually I'm going to open a new PR with the docstring and test changes that I would like to salvage.