Support multi-residue molecules in template generator and improve performance

[epretti]

• Support multi-residue molecules, including but not limited to proteins, in template generators (resolves SMIRNOFFTempalteGenerator and proteins #415)
• Uses OpenMM's template matcher instead of NetworkX, and look up molecules by chemical formulas before trying template matching (resolves TemplateGenerator._match_residue() is too slow #426, and openmmforcefields can automatically take advantage of any future performance improvements to OpenMM template matching)
• Changes template generator cache behavior so that molecules must be specified explicitly in order to be parameterized (resolves Unexpected / unintuitive behavior of template generator caches #429)

Todo:

• Add test including multiple residues and virtual sites
• Fix CHARMM force fields updated by Templates can match whole molecules openmm#5181

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.73%. Comparing base (0551fd1) to head (8eb86a5).
⚠️ Report is 1 commits behind head on main.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #430      +/-   ##
==========================================
+ Coverage   83.70%   84.73%   +1.02%     
==========================================
  Files           5        5              
  Lines         798      799       +1     
==========================================
+ Hits          668      677       +9     
+ Misses        130      122       -8     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[j-wags]

I've been stress-testing this extensively over the past few days and haven't found any significant issues. Details of my testing are here. Other than yielding an issue with the new constraint functionality in OpenMM this behaves well in every situation I can think of. Thanks for all the hard work and attention to detail, @epretti! This is good to merge as far as I can tell.

[epretti]

Thanks for the detailed look at this. If no other concerns come up between now and our check-in tomorrow, I'll plan on merging this afterwards.

[mattwthompson]

No concerns from me

[jthorton]

toolkit molecules also have a to_hill_formula() method incase that is prefered.

Suggested change

	formula = tuple(sorted(collections.Counter(atom.atomic_number for atom in molecule.atoms).items()))
	formula = molecule.to_hill_formula()

[jthorton]

Ah never mind you need it to be consistent with the formula method below for the openmm residue.

[epretti]

This should be ready to merge.

[epretti]

As usual, I don't have the permissions to merge in this repository, so someone else will have to be responsible for pressing the approve button :)

[peastman]

I'm not quite sure why it doesn't let you. I tried to change the settings, and that seems to have caused it to be merged automatically. Let's see what happens on the next PR.

[epretti]

OK. From what I recall in the past, the repository was set up so that PRs could not be merged without "one approving review from someone with write access". And (as makes sense when this feature is enabled) you cannot approve your own PR.

[peastman]

That makes sense. I'm ok with not requiring an approval if you are.

[epretti]

I think it's fine if you want to disable the check (since we don't have it on the main repository). I will always wait for approval before merging anyway; it's just that sometimes this happens in a comment or a conversation.