Set minimum for openmmforcefields version in conda envs to 0.12.0

[dotsdl]

Even though 0.12.0 is currently the latest version of openmmforcefields, building the env without forcing it appears to get 0.11.2 at the moment.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (813cc5a) 82.85% compared to head (36efbda) 82.26%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #193      +/-   ##
==========================================
- Coverage   82.85%   82.26%   -0.59%     
==========================================
  Files          22       22              
  Lines        2735     2735              
==========================================
- Hits         2266     2250      -16     
- Misses        469      485      +16     

see 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[dotsdl]

Looks like ambertools is keeping us from using python 3.11? @mikemhenry any insights as to why this may suddenly be the case with openmmtools 0.12.0?

I'm fine with downgrading to Python 3.10 in this instance, but wanted to raise this as odd in case it's the first we've seen it.

[mikemhenry]

We don't have python 3.11 builds for older ambertools, and we can't use the newest openmmtools with the older ambertools.

Does that make sense? In a perfect world, I would do a amber22 build for python 3.11

[dotsdl]

@mikemhenry yes, that makes sense. As discussed today, can you take a stab at getting an ambertools 22 build for Python 3.11, since I think without that, practically the whole openforcefield stack and everything using it will be held back to Python 3.10. Not exactly urgent, but seems unlikely to resolve on its own?

For (my future) reference, here is the pin that holds us back from ambertools 23: https://github.com/conda-forge/openmmforcefields-feedstock/blob/543afe9bd8fc8e6dbc3084ba59d0bfeaa26a358b/recipe/meta.yaml#L25

[dotsdl]

For the moment then here, we will remove Python 3.11 as a tested version, and proceed with deploying on Python 3.10. The CI for that version will always fail at present given the problem above.

[dotsdl]

Decided to keep 3.11 CI test; we don't require it to pass before merge, so this keeps it around as a reminder that this needs fixing. May revisit this decision if it proves annoying having it around.