Skip unit tests for docs-only changes

[nbiederbeck]

Shall we try to reduce the CI time for documentation-only changes? I think of running just the docs CI when having the label documentation attached to the issue, and ignoring the whole pytest suite.

That would mean that we need to split our current pytest approach, where we test "normal" tests and documentation tests simultaneously.

This would greatly reduce the time to passing tests in minor (or large) documentation changes, and we would not be held up by increased memory usage, suddenly incompatible dependencies etc. like we were in the last few days.

(I'm thinking in particular of PRs like #2220)

[kosack]

Yes that would be fine - changes to the docs only should not require pytest to be run. And maybe it would encourage people to work more on the docs as well.