Fix issue with sampling of PartialObservedRVs

[ricardoV94]

Description

There was issue with PartialObservedRVs in that we used the same RNG in two places: in one of the masked RVs and in the shape of the empty tensor where the two masked RVs are stored as a deterministic.

This would further raise an error in JAX when computing the joined deterministic, because we don't allow RNGs in jaxified graphs.

Constant folding will ensure the shape graph will be introduced instead of the default ShapeOp(RV) so the RNG shouldn't be part of that deterministic graph.

Related Issue

Issue reported in https://discourse.pymc.io/t/shared-randomtype-issue-using-nuts-numpyro-w-r-t-data-containing-nans/13503?u=ricardov94

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://pymc--7071.org.readthedocs.build/en/7071/

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (35cd657) 92.20% compared to head (24527c2) 92.20%.
Report is 3 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #7071   +/-   ##
=======================================
  Coverage   92.20%   92.20%           
=======================================
  Files         101      101           
  Lines       16895    16901    +6     
=======================================
+ Hits        15578    15584    +6     
  Misses       1317     1317           

Files	Coverage Δ
pymc/distributions/distribution.py	96.26% <100.00%> (+<0.01%)	⬆️
pymc/distributions/multivariate.py	93.53% <100.00%> (+0.04%)	⬆️
pymc/logprob/transform_value.py	93.75% <100.00%> (ø)
pymc/pytensorf.py	91.26% <100.00%> (-0.03%)	⬇️