How should our test matrix look like?

There have been a few conversations about how our test matrix should look like. There are clearly costs and benefits to all configurations and I think we should settle what our strategy is to not have individual discussions on tickets

### Related tickets
- https://github.com/dask/distributed/pull/6073
- https://github.com/dask/distributed/pull/5648
- https://github.com/dask/distributed/pull/4794

----

One proposal

Currently we have:

Python version | OS | Which deps | deps version
--- | --- | --- | ---
3.8 | Linux, MacOSX, Windows | mandatory, optional, Cython | latest
3.9 | Linux, MacOSX, Windows | mandatory, optional, CUDA | latest
3.10 | Linux, MacOSX, Windows | mandatory, optional | latest+git tip

I think we should change it to:

Python version | OS | Which deps | deps version
--- | --- | --- | ---
3.8 | Linux | mandatory, optional, CUDA | pinned to minimum
3.8 | Linux | mandatory only (no numpy!) | latest
3.8 | Linux, MacOSX, Windows | mandatory, optional, CUDA | latest
3.9 | Linux | mandatory, optional, CUDA | latest
3.10 | Linux, MacOSX, Windows | mandatory, optional, CUDA | latest+git tip

That's 9 workflows before and after, but with one less Mac workflow and *a lot* of added value.
I expect the first two workflows to require a lot of effort before they become green.

_Originally posted by @crusaderky in https://github.com/dask/distributed/issues/6073#issuecomment-1091765866_