Audit legacy edge-case warning semantics

[bernalde]

Context

While following up on PR #22, the CI-equivalent non-notebook/non-Pyomo test subset originally reported 2064 warnings. PR #22 now filters the expected test warnings narrowly in pytest.ini, so the test run is clean again.

The warnings came from legacy edge-case tests that intentionally exercise infeasible or no-sublimation paths:

• repeated No sublimation... UserWarnings from lyopronto.calc_unknownRp
• Total shelf temperature setpoint time exceeded... from lyopronto.calc_unknownRp
• infeasible design-space warnings from lyopronto.design_space
• pressure/shelf-temperature infeasibility warnings from lyopronto.calc_knownRp
• optimization failure and incomplete-drying warnings from lyopronto.opt_Pch
• a small number of RuntimeWarning: divide by zero encountered in scalar divide from lyopronto.functions.Rp_finder

Why this should be separate from PR #22

The warning flood is test noise, but the underlying code behavior is broader than benchmark validation/timeouts. In particular, Rp_finder has a singular denominator when Tbot == T_sub; the current downstream behavior appears tolerated by tests, but changing it should be reviewed against physical/numerical expectations.

Suggested Work

• Decide whether these conditions should use custom warning categories rather than plain UserWarning.
• Audit Rp_finder behavior for Tbot == T_sub and document or handle the singular resistance case explicitly.
• Consider adding a verbosity or diagnostics option for calculators that can emit per-time-step warnings.
• Replace broad coverage-test warning filters with pytest.warns at specific call sites where practical.

Acceptance Criteria

• Expected infeasible/no-sublimation scenarios are still visible to users when appropriate.
• Tests do not emit large warning counts during normal CI runs.
• Any singular resistance behavior is documented and covered by focused tests.

[bernalde]

Closed via #25