Audit edge-case warning semantics

[bernalde]

Summary

• Removed the legacy repository-wide warning filters from pytest.ini.
• Made lyopronto.functions.Rp_finder handle the exact Tbot == T_sub singular case explicitly.
• Replaced broad warning suppression with focused pytest.warns or test-local pytest.mark.filterwarnings in the edge-case tests that intentionally exercise infeasible/no-sublimation paths.

Warning Filters

• Removed all global filters for calc_unknownRp, calc_knownRp, design_space, opt_Pch, and the Rp_finder divide-by-zero RuntimeWarning from pytest.ini.
• Added module-local filters in tests/test_calc_unknownRp_coverage.py for repeated expected No sublimation and setpoint-exhaustion warnings from coverage-only unknown-Rp scenarios.
• Added test-local filters in tests/test_coverage_gaps.py for expected design-space infeasibility and single-timestep warnings.
• Added test-local filters in opt_Pch coverage/comparison tests where optimizer failure or incomplete drying is the expected tolerated edge-case behavior.
• Changed tests/test_calculators.py::TestEdgeCases::test_very_low_shelf_temperature to assert the expected known-Rp vapor-pressure warning with pytest.warns.
• Kept no repository-wide warning filters. No custom warning categories were introduced.

Rp_finder

Rp_finder now treats a zero bottom-to-sublimation temperature gradient as a documented singular resistance case instead of relying on NumPy divide warnings:

• nonzero pressure driving force returns signed infinity;
• zero pressure driving force returns NaN for the indeterminate 0/0 case;
• focused tests assert the behavior and confirm no warning is emitted.

Tests Run

• pytest tests/test_functions.py tests/test_calc_unknownRp.py tests/test_calc_knownRp.py tests/test_design_space.py tests/test_opt_Pch.py tests/test_calc_unknownRp_coverage.py tests/test_coverage_gaps.py -q -o addopts= -o filterwarnings=error --maxfail=10
  • Result: 127 passed in 180.20s
• python -m pytest tests/test_calculators.py::TestEdgeCases::test_very_low_shelf_temperature tests/test_opt_Pch_coverage.py tests/test_opt_Pch_Tsh.py::TestOptPchTshValidation::test_joint_optimization_faster_than_single tests/test_opt_Pch_Tsh_coverage.py::TestOptPchTshComparison::test_joint_opt_vs_pch_only -q -o addopts= -o filterwarnings=error
  • Result: 16 passed in 111.95s
• python -m pytest tests/ -n auto -q -m "not notebook and not pyomo" --ignore=tests/test_pyomo_models
  • Result: 257 passed in 359.27s
• git diff --check
  • Result: passed
• python -m mypy lyopronto/functions.py
  • Result: Success: no issues found in 1 source file
  • Note: local mypy also prints that configured Python 3.8 is no longer supported by this installed mypy version.

Notes

• python -m ruff check on the changed files was run and reports existing legacy style/import/trailing-whitespace issues across these files; those are outside this warning-semantics change.
• python -m black --check on the changed files was run and reports that the legacy files would be reformatted; no formatting pass was applied to avoid unrelated churn.
• Notebook and Pyomo solver tests were not run locally; the requested practical non-notebook/non-Pyomo subset was run.

Closes #23

[bernalde]

Reviewed against issue #23 and the surrounding warning-producing paths. I did not find blocking issues. The PR removes the repository-wide warning filters, keeps expected noisy warnings localized to tests that intentionally exercise those paths, and adds focused coverage for the Rp_finder Tbot == T_sub singular cases. Targeted warning-as-error tests and the practical non-notebook/non-Pyomo test subset pass locally. Merge-ready from my review; GitHub would not allow an APPROVE review from the PR author account.