Add transition-state thermochemistry and Eyring rate constants

[galjos]

Thermo.run() previously raised on any imaginary vibrational frequency, so a transition state (a first-order saddle point) could not be evaluated at all — blocking barrier/rate work.

Core: Thermo(..., transition_state=True)

• Requires exactly one imaginary mode among the kept frequencies (raises a clear error otherwise: too few/too many).
• Excludes that mode from the vibrational partition function / entropy / energy / heat-capacity sums (the reaction coordinate contributes no thermal vibrational term) — everything else (rotational/translational/electronic) is unchanged.
• Exposes it via Thermo.imaginary_mode_wavenumber() (cm⁻¹, negative).
• Non-TS behavior (the default) is byte-for-byte unchanged. temperature_scan propagates the flag.

Plumbing

transition_state=False threaded through run_thermo, orca_thermo, cclib_thermo, pyscf_thermo — the "import an externally-computed structure" engines, since a DFTB+/xtb geometry optimization is a minimizer and can't itself land on a saddle point.

New: ThermoScreening.thermo.kinetics

• eyring_rate_constant(reactants, ts, temperature, kappa=1.0) — TST rate constant k = kappa * (kB T/h) * exp(-ΔG‡/RT), using the same Thermo/(coefficient, Thermo) stoichiometry convention as reaction_free_energy.
• wigner_tunneling_correction(imaginary_wavenumber, temperature) — a simple first-order tunneling estimate from the TS's imaginary frequency.

Verification

• Manually verified k(ΔG‡=0) = kB*T/h (~6.212×10¹² s⁻¹ at 298.15 K, the exact textbook TST prefactor) and cross-checked a 20 kcal/mol barrier against the well-known "ΔG‡≈17.5 kcal/mol → k≈1 s⁻¹" rule of thumb — both match exactly.
• New tests for the core TS logic (exclude-one-mode, wrong-count errors, temperature_scan propagation), kinetics.py (closed-form cross-checks, stoichiometry, kappa scaling), and one transition-state forwarding test per engine (run_thermo, orca_thermo, cclib_thermo, pyscf_thermo).
• Full suite: 365 passed. Docs updated (usage + API reference). Full patch coverage (verified diff-line-by-line, not just file-level).

Closes #104

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.25%. Comparing base (88f3db1) to head (3fdf181).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #105      +/-   ##
==========================================
+ Coverage   97.22%   97.25%   +0.03%     
==========================================
  Files          30       31       +1     
  Lines        1872     1895      +23     
==========================================
+ Hits         1820     1843      +23     
  Misses         52       52              

Flag	Coverage Δ
unittests	97.25% <100.00%> (+0.03%)	⬆️

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.