Handle lower flag more carefully in Solve to CholeskySolve rewrite

[jessegrabowski]

Description

There was some discussion in #1467 about whether the reuse_decomposition_multiple_solves rewrite was doing the right thing with the lower flags, especially in the corner case where a user is being really clever and only storing half the symmetric matrix.

This PR handles the case more carefully, and checks on a matrix with NaN values that we don't look at the wrong triangles after rewriting.

It also forwards the check_finite argument in the Cholesky Op, which was not being respected previously. I think this is also being fixed in #1487 , so there's some undesirable overlap between these PRs.

Related Issue

• Closes #
• Related to Reuse cholesky decomposition with cho_solve in graphs with multiple pt.solve when assume_a = "pos" #1467

Checklist

• Checked that the pre-commit linting/style checks pass
• Included tests that prove the fix is effective or that the new feature works
• Added necessary documentation (docstrings and/or example notebooks)
• If you are a pro: each commit corresponds to a relevant logical change

Type of change

• New feature / enhancement
• Bug fix
• Documentation
• Maintenance
• Other (please specify):

---

📚 Documentation preview 📚: https://pytensor--1492.org.readthedocs.build/en/1492/

[jessegrabowski]

I reset the logic of the rewrite in this PR, but added a test of behavior under different lower flags. I think we were doing the right thing from the get-go, but we might have to think about this a bit more carefully.

Even if everything is fine, I think having this new test is good.

[ricardoV94]

Shall we close this?

[jessegrabowski]

Yes, we resolved it (or maybe it was never a problem)