Virtual site overhaul

[mattwthompson]

Description

Resolves #789

Checklist

• Fix BondCharge positions
• Test OpenMM weights with a BondCharge placed at various locations
• Use TwoParticleAverageSite for BondCharge
• Use ThreeParticleAverageSite for some MonovalentLonePair
• Test positions of TrivalentLonePair
• Test geometries of four-site water models
• Use OutOfPlaneSite for DivalentLonePair
• Implement MonovalentLonePair positions
• Implement DivalentLonePair positions
• Implement TriivalentLonePair positions
• Compare TIP4P implementations to OpenMM
• Compare TIP5P implementations to OpenMM
• Add tests
• Lint
• Update docstrings

[codecov]

Codecov Report

Merging #791 (084be46) into main (fdf7630) will decrease coverage by 0.34%.
Report is 3 commits behind head on main.
The diff coverage is 96.80%.

❗ Current head 084be46 differs from pull request most recent head 6cbdea0. Consider uploading reports for the commit 6cbdea0 to get more accurate results

Additional details and impacted files

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[mattwthompson]

I'm confused as to how the ligand + water example is now failing; of the things I expected to break, atom indexing wasn't one of them. I spent too long on this tonight, will pick it up again tomorrow.

[mattwthompson]

The math of TrivalentLonePair is trickier but I think I have things right; at very least the direction appears to be consistent with the SMIRNOFF spec (positive: "above" the nitrogen; negative: "below" the nitrogen) and qualitatively matching the value of the parameter. Here are visualizations with distance values of -0.5 A and +2.5 A:

[mattwthompson]

There are a couple of things left to do but I'm going to put this out anyway. This fixes a fairly important bug and putting a release out might also fix #790