Recognize affine functions of independent Gaussians as MvNormal logp

[ricardoV94]

Teach pymc.logprob to derive the logp of a value variable that is an
affine function of one or more independent Gaussian leaves (e.g. the
low-rank guide loc + W @ ek + d * ed) by propagating moments and
substituting a dense MvNormal, instead of inverting the transform.

This unlocks fundamentally non-invertible cases the existing machinery
declines: wide/low-rank matvecs (find_measurable_matmul bails on
non-square matrices) and sums of >=2 independent Gaussians (the elemwise
transforms require <=1 measurable operand). Covers low-rank ADVI guides,
sums of independent normals, and linear-Gaussian marginalization
(regression coefficients, etc.).

The accumulated covariance is kept dense; exploiting low-rank structure
for an O(D K^2) logp would need a Woodbury rewrite in PyTensor, which
does not exist yet, so the emitted MvNormal does a dense O(D^3) cholesky.

Claude-Session: https://claude.ai/code/session_0183HnSTpYk7UrvAJ7D3kBtj

[read-the-docs-community]

Documentation build overview

📚 pymc | 🛠️ Build #33330413 | 📁 Comparing b22c382 against latest (558bb5e)

  🔍 Preview build  

1 file changed

± glossary.html

[codecov]

Codecov Report

❌ Patch coverage is 85.48387% with 27 lines in your changes missing coverage. Please review.
✅ Project coverage is 86.71%. Comparing base (fffac80) to head (b22c382).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
pymc/logprob/gaussian.py	85.40%	27 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8342      +/-   ##
==========================================
- Coverage   86.73%   86.71%   -0.03%     
==========================================
  Files         126      127       +1     
  Lines       20595    20781     +186     
==========================================
+ Hits        17863    18020     +157     
- Misses       2732     2761      +29     

Files with missing lines	Coverage Δ
pymc/logprob/__init__.py	100.00% <100.00%> (ø)
pymc/logprob/gaussian.py	85.40% <85.40%> (ø)

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

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