Fix zarr append with groups

[niowniow]

Fixes the issue that xarray.core.dataset.Dataset.to_zarr produced errors when using append_dim and group simultaneously. The issue was that during appending, the zarr store was opened without the group information.

• ClosesDataset.to_zarr() with mode='a' does not work with groups #3170
• Tests added
• Passes black . && mypy . && flake8
• Fully documented, including whats-new.rst for all changes and api.rst for new API

[dcherian]

Thanks @niowniow.

This looks OK to me but I don't know the zarr code.

[dcherian]

@niowniow can you merge master please? The diff is huge right now

[niowniow]

I tried to merge with master yesterday and did it again today. There might be a problem with the way I tried to fix the whats-new.rst issues. It might be faster if I open a new pull request and re-insert the changes!?

[keewis]

If you don't mind I'll try to fix that for you.

[niowniow]

I don't mind, thanks :)

[keewis]

I cherry-picked your commits, but I'm not sure I included all the changes. Could you check that?

Also, I can't seem to figure out how to push to this PR, so you will have to do that instead:

git checkout fix_zarr_append_with_groups
# backup branch
git branch backup_fix_zarr_append_with_groups
# delete branch
git checkout master
git branch -D fix_zarr_append_with_groups
# checkout fixed branch
git checkout --track origin/fix_zarr_append_with_groups
# force push
git push -f <fork-remote> fix_zarr_append_with_groups

[niowniow]

Okay, this looks better. Thanks for the help! I had to try it again myself, because you cherry-picked some commits I accidentally merged into this branch (non-related to this fix).
Could you check if this is good now?

[keewis]

Looks good to me. However, I don't know much about backends so I'll let someone else review the code.

[rabernat]

Sorry for my very slow review of this. I just looked over the PR and found zero issues. Thanks for your patience.

[rabernat]

@dcherian - I assume these test failures (py38-upstream-dev and docs) are acceptable? They seem unrelated to the PR.

If so, please merge.

[keewis]

the upstream-dev failure is #3751, but although there should be no change here that would trigger it, the docs failure (referencing api using :ref: instead of :doc:) does not seem to happen with other PRs (I also can't reproduce it locally).

[rabernat]

the docs failure (referencing api using :ref: instead of :doc:) does not seem to happen with other PRs (I also can't reproduce it locally

I can't find that message here: https://dev.azure.com/xarray/xarray/_build/results?buildId=2214&view=logs&jobId=7e620c85-24a8-5ffa-8b1f-642bc9b1fc36&j=7e620c85-24a8-5ffa-8b1f-642bc9b1fc36&t=68484831-0a19-5145-bfe9-6309e5f7691d

To me it looks like the doc build failed due to two warnings: failure to fetch the pandas intersphinx obj and a sphinx deprecation warning about scripts.

[keewis]

it's here:

/home/vsts/work/1/s/doc/faq.rst:239: WARNING: undefined label: api (if the link has no caption the label must precede a section header)

not sure if the intersphinx warning also makes the build fail, but that could be fixed by rerunning. The deprecation warning is not a problem.

[dcherian]

All tests are green. Thanks @niowniow. I see this is your first contribution. Welcome!