Testing different executors automatically

[TomNicholas]

We briefly discussed the difficulty of maintaining multiple executors. In https://github.com/tomwhite/cubed/pull/168#issuecomment-1542892933 I suggested expanding the CI to run different test jobs with different executors installed.

I also saw this dask blog post about maintaining compatibility with changes to multiple upstream libraries, and they shoutout this GH action that creates an issue automatically when an upstream-dev CI build fails, which we use in xarray that's very useful.

None of this is particularly pressing but might be useful later.

EDIT: More info about all of dask's automated performance testing that might be useful to come back to.