test: fix shear-wave review findings (runner finiteness, source pin, contract)
- run_case: finite_row accepts list diagnostics; sample() shared by sample
and READY paths; verify_source pins shear_wave.py path+SHA256 for
--flow shear_wave (default TG flow unchanged, does not import it).
- assess_shear_wave.py: fixed N128 four-case contract and 4-JSON CLI,
reusing assess_projected_pairs provenance/build/workload/time checks plus
shear metadata; missing/duplicate/mixed/extra records -> INVALID_SET.
- tests: mocked refined runner sample (list diagnostics, NaN, TG unchanged),
contract pass + reproduced false-pass mutations, divergence check uses
analytic E.k=0 and a derived h^2 truncation + roundoff bound with
h-halving scaling instead of a fixed 1e-12.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
