Bug fix: dimers + unique_pairs_only

[chrisiacovella]

Pull Request Summary

This fixes a bug where the shape of the underlying tensors in the PairListData was incorrect when evaluating the neighbors of a dimer with unique_pairs_only.

This issue stemmed from squeezing the array for masking; when that mask only had dimension 1, it was squashed. The addition of a reshape(-1) fixes this issue. Tests for the dimer edge case were also added.

This also addresses the issue with calculating the AngularSymmetryFunction for dimers. As discussed in the linked issue, for a dimer, where there are no angles, this tensor is empty. Reshaping the empty tensor was undefined; changing this to use flatten instead works for all cases. This edge case was missed in the ani testing due to the fact the 1000 conformer test say does not feature one of the ~10 dimer systems in the dataset (and thus failure was only observed when training with the full dataset). Additional tests still are required to cover this edge case.

Associated Issue(s)

• ANI2x training failure #347

Pull Request Checklist

• Issue(s) raised/addressed and linked
• Includes appropriate unit test(s)
• Appropriate docstring(s) added/updated
• Appropriate .rst doc file(s) added/updated
• PR is ready for review

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 81.58%. Comparing base (ba16705) to head (25395b7).
⚠️ Report is 382 commits behind head on main.

❌ Your project status has failed because the head coverage (81.58%) is below the target coverage (90.00%). You can increase the head coverage or adjust the target coverage.

🚀 New features to boost your workflow:

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