Drop divergence_message in drop_warning_stat

[gaoflow]

Description

nutpie can add a string-valued divergence_message variable to sample_stats, which keeps keep_warning_stat=False from producing an InferenceData object that can be serialized cleanly.

This updates drop_warning_stat() to drop divergence_message alongside the existing warning stats, while preserving ordinary sampler stats like diverging. A regression test covers both sample_stats and warmup_sample_stats.

AI assistance was used under my direction.

Related Issue

• Closes BUG: keep_warning_stat argument in pm.sample does nothing #8321
• Related to #

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):

Local checks

• uv run --with-editable . --with pytest python -m pytest tests/test_util.py::test_drop_warning_stat tests/test_util.py::test_drop_divergence_message_stat -q
• uv run --with ruff ruff check pymc/util.py tests/test_util.py
• uv run --with ruff ruff format --check pymc/util.py tests/test_util.py
• git diff --check origin/main...HEAD

[read-the-docs-community]

Documentation build overview

📚 pymc | 🛠️ Build #33218546 | 📁 Comparing 2d098fd against latest (9d24260)

  🔍 Preview build  

58 files changed · ± 58 modified

± Modified

• glossary.html
• api/dims/generated/pymc.dims.Beta.html
• api/dims/generated/pymc.dims.Cauchy.html
• api/dims/generated/pymc.dims.Gamma.html
• api/dims/generated/pymc.dims.HalfCauchy.html
• api/dims/generated/pymc.dims.HalfNormal.html
• api/dims/generated/pymc.dims.HalfStudentT.html
• api/dims/generated/pymc.dims.InverseGamma.html
• api/dims/generated/pymc.dims.Laplace.html
• api/dims/generated/pymc.dims.LogNormal.html
• and 48 more...

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.70%. Comparing base (9d24260) to head (2d098fd).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #8334       +/-   ##
===========================================
+ Coverage   79.80%   91.70%   +11.90%     
===========================================
  Files         125      125               
  Lines       20526    20526               
===========================================
+ Hits        16380    18824     +2444     
+ Misses       4146     1702     -2444     

Files with missing lines	Coverage Δ
pymc/util.py	81.09% <ø> (+2.94%)	⬆️

... and 42 files with indirect coverage changes

🚀 New features to boost your workflow:

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

[gaoflow]

CI triage note: the current red job looks unrelated to this drop_warning_stat change.

The failing matrix is alternative_backends (cvm, 3.13, tests/distributions/test_random_alternative_backends.py ...), and the only assertion failure shown in the log is:

FAILED tests/distributions/test_random_alternative_backends.py::test_mvstudentt[NUMBA]
Not equal to tolerance rtol=0.1, atol=0
ACTUAL: array([0.999302, 2.002764, 2.680959])
DESIRED: array([1, 2, 3])

The all_tests check then fails only as the aggregate matrix-status step. This PR only changes pymc/util.py and tests/test_util.py to drop the string-valued divergence_message sample stat, so I do not see a code action for this branch from that failure.