Close validation gaps: broaden cross-checks, benchmark, and document

[galjos]

Follow-up to the end-to-end validation (which found no bugs) — turns each remaining coverage gap into a durable artifact. No behaviour changes: every source edit is a docstring. Every reference number was produced by a real calculation and adversarially re-verified.

CI-safe tests (no backend)

• test_heat_capacity_cv.py (new) — settles the Cv-vs-Cp question: the reported heat capacity is the ideal-gas constant-volume Cv, cross-checked non-circularly against the closed-form Cv and pinned values (also confirmed tool Cv == native-xtb Cp − R).
• test_S_H_G_match_ase — broadens the ASE IdealGasThermo cross-check from entropy-only/6 molecules to S + H + G across 14 (incl. charged NH₄⁺/OH⁻ and larger benzene/ethane/methanol), masses pinned, agreement < 2e-3.
• test_quasi_rrho_matches_independent_grimme_and_is_invariant — qRRHO vs an independent hand-coded Grimme formula; confirms enthalpy and Cv are invariant under quasi_rrho.
• test_dftb_setup.py — solvent→GBSA param plumbing for water/dmso/acetonitrile/thf.

Skippable integration tests (need dftb+/tblite)

• xTB and DFTB+ gas-phase entropy vs experiment (H₂O/CH₄/N₂, within 2 cal/mol/K) — closes the "no experimental accuracy" gap.
• qRRHO end-to-end on a real floppy molecule (n-octane, embedded geometry) with engine-generated sub-100 cm⁻¹ modes (28/59/71) → ΔS ≈ −4 cal/mol/K.
• same-machine reproducibility for both engines.
• DFTB+ solvation across dmso/thf/water: GBSA block applied, dmso/thf stabilise, and the lowest frequency shifts (solvation enters the Hessian).

Docs

• total_heat_capacity now documents Cv vs Cp; dftbplus_thermo/xtb_thermo get reproducibility + solvation-accuracy notes; Thermo pressure unit corrected to Pa.
• docs/benchmarks/entropy_accuracy.md records the measured accuracy vs experiment (xTB MAD 0.17, DFTB+ 3ob MAD 0.14 cal/mol/K).

Validation

Offline suite 241 passed / 10 skipped; with real DFTB+ + tblite 251 passed / 0 skipped. Source changes are docstring-only, so codecov/patch has no new coverable lines.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 96.57%. Comparing base (fd6cac2) to head (724ae84).

Additional details and impacted files

@@           Coverage Diff           @@
##             main      #69   +/-   ##
=======================================
  Coverage   96.57%   96.57%           
=======================================
  Files          23       23           
  Lines        1459     1459           
=======================================
  Hits         1409     1409           
  Misses         50       50           

Flag	Coverage Δ
unittests	96.57% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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