Eagerly clean rv shape references introduced by maybe_resize

[ricardoV94]

Closes #8301

Size arguments don't make it into regular logprob expressions, so we monkey patch that with maybe_resize in #8105

Otherwise it's fine for arguments to include shape(rv) in their arguments, PyTensor conditional_logprob will replace those with references to the shape(value) before the logprob function is invoked. TruncatedRV however can refer the shape of a variable that is not in the graph (the generating untruncated dist) and introduce that in a way that conditional_logprob wouldn't clean.

This PR is not the ultimate answer to this question. First the current check could be said to be too eager, RV in the graph just inside the shape argument are plausible fine. Or we could always try to call resolve_shape before failing. Still seems like a good place to do this for now, we can iterate later if it is not good enough.

The new helpers have long been needed and reinvented in multiple places, so it's good to have them regardless.

[read-the-docs-community]

Documentation build overview

📚 pymc | 🛠️ Build #32735813 | 📁 Comparing 5faa634 against latest (a65e8b6)

  🔍 Preview build  

4 files changed

± glossary.html
± contributing/jupyter_style.html
± _modules/pymc/pytensorf.html
± _modules/pymc/distributions/shape_utils.html

[codecov]

Codecov Report

❌ Patch coverage is 90.90909% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 91.77%. Comparing base (169e901) to head (5faa634).

Files with missing lines	Patch %	Lines
pymc/pytensorf.py	87.50%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8302      +/-   ##
==========================================
- Coverage   91.78%   91.77%   -0.01%     
==========================================
  Files         125      125              
  Lines       20428    20437       +9     
==========================================
+ Hits        18749    18757       +8     
- Misses       1679     1680       +1     

Files with missing lines	Coverage Δ
pymc/distributions/shape_utils.py	92.38% <100.00%> (+0.07%)	⬆️
pymc/pytensorf.py	88.44% <87.50%> (-0.05%)	⬇️

🚀 New features to boost your workflow:

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

[ricardoV94]

@maresb can you check if this fixes the actual issue you had and not just the MWE?

[maresb]

Wow, thanks so much!!!

Will check in the morning.

[maresb]

Thanks so much @ricardoV94! Confirmed that this resolves also my non-minimal example.