Raise when reduced dims are chunked in map_blocks

[aulemahal]

Pull Request Checklist:

• This PR addresses an already opened issue (for bug fixes / features)
  • This PR fixes Odd behaviour when chunks are used along quantiles and time grouper dimensions in SDBA EQM. #1481
• Tests for the changes have been added (for bug fixes / features)
  • (If applicable) Documentation has been added / updated (for bug fixes / features)
• CHANGES.rst has been updated (with summary of main changes)
  • Link to issue (:issue:number) and pull request (:pull:number) has been added

What kind of change does this PR introduce?

If a map_blocks-wrapped function receives input chunked along the dimensions marked as "reduced", it raises an error.

This means one cannot chunk the training dataset (QM.ds, for example) along the distribution ('quantiles') or group ('month', 'dayofyear') dimensions.

Previously, this didn't raise an error, but the adjustment would be done separately for each chunk, yielding incorrect results, silently.

Does this PR introduce a breaking change?

No.

Other information:

[github-actions]

Warning
This Pull Request is coming from a fork and must be manually tagged approved in order to perform additional testing.

[aulemahal]

@github-actions, this ain't a fork, this is the OG xclim! Bad bot.

[Zeitsperre]

@aulemahal

I wish the bot was at fault here, haha. Will push a fix soon.

[coveralls]

coverage: 90.492% (+0.02%) from 90.468% when pulling c846ca6 on fix-1481 into 0a48bbc on master.