Fix FusionOptimizer bug

[ricardoV94]

First commit is just small refactor, second fixes the issue.

Reported by @jessegrabowski, where pymc model would hang before sampling. We need to update the ancestors_bitset when we implicitly "claim" a subgraph.

[jessegrabowski]

I approved too fast, it looks like you broke scan :D

[ricardoV94]

I approved too fast, it looks like you broke scan :D

Oh well we don't lose much anyway

[codecov]

Codecov Report

❌ Patch coverage is 94.11765% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 81.67%. Comparing base (b05acfd) to head (ca797da).
⚠️ Report is 4 commits behind head on main.

Files with missing lines	Patch %	Lines
pytensor/tensor/rewriting/elemwise.py	94.11%	1 Missing ⚠️

❌ Your patch status has failed because the patch coverage (94.11%) is below the target coverage (100.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1630      +/-   ##
==========================================
- Coverage   81.67%   81.67%   -0.01%     
==========================================
  Files         232      232              
  Lines       53132    53132              
  Branches     9410     9407       -3     
==========================================
- Hits        43396    43395       -1     
- Misses       7283     7284       +1     
  Partials     2453     2453              

Files with missing lines	Coverage Δ
pytensor/tensor/rewriting/elemwise.py	93.46% <94.11%> (-0.17%)	⬇️

🚀 New features to boost your workflow:

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

[ricardoV94]

good news: it works now. bad news: mypy ...

[ricardoV94]

this was what was causing the scan tests to fail. If we add the subgraph_bitset of failed nodes to all_subgraphs_bitset it then causes a failure in the sorting of the subgraphs below, because it was trying to sort wrt to these ones that don't matter.

[jessegrabowski]

What was the original logic of including it? Or was it just a mistake

[ricardoV94]

original this variable was meant more as a collection of "variable that were claimed or known to not be fuseable", so it would fulfill the second goal. But that info was not used anywhere in the latest iteration. Instead we only use this for the sorting bit at the end.

[ricardoV94]

This would have raised in the failing scan test, showing our sorting failed