Tests fail due to pyiron_atomistics and scikit_image conflict

[liamhuber]

The latest version of pyiron_atomistics wants scipy>=1.10.1, but scikit-image v0.20.0 has a lower bound of >=1.8 and a conditional upper bound of <1.9.2. Interestingly, an informative error message only appears on Raynol's PR, the rest just crash out before printing the important information.

The good news is that scikit-image now has 0.21.0, which removes the upper bound for scipy. The bad news is, this is not yet released on conda forge, nor does their feedstock yet have a PR for this version.

Still, as soon as they release their latest version on conda, all we will need to do is bump our dependency and this should resolve itself.

@niklassiemer, @raynol-dsouza, I'll try to check back daily on this and make the dependency bump ASAP.
Ahhh, shoot. I was just looking to see what the time lag was between their github and conda version releases and I see now that 0.21.0 is actually still just the "pre-release" 0.21.0rc0... So there is no real concrete timeframe for when they're going to release the version that will fix our problem.

I'm going to make a PR pinning us to an older version of scikit-image -- the next-most-recent, 0.19.3, only pins scipy to >=1.4.1. Their dependency pins are a bit complex and python-version-dependent, so I'm not sure whether this will introduce some other dependency clash, but let's try it. Then we can keep going with our own work and skip straight from 0.19 to 0.21 whenever the latter eventually gets released.