Document nonportable RM3 End Stops work audit

[cmccomb]

Result

The independently advanced public WEC.run(..., solver="wecsim_ode45") RM3 End Stops case gives a 400 s total damper-work gap of 0.4307% on local ARM64 macOS, below the existing 0.7% limit. Two local 100 s windows already exceed that limit (0.7993% and 0.9021%). On the exact prior PR head, the full Linux x86 Python suite measured a 0.8246% total-work gap and failed the unchanged 0.7% gate in run 39369510668. Both environments used NumPy 2.5.4 and SciPy 1.18.1.

This revision removes the attempted portable full-run gate and records the counterexample in PARITY.md and the case-status ledger. The existing independent 120 s motion gate remains. RM3 End Stops stays qualified; badges and the physical solver default do not change. No threshold was relaxed.

Verification

• Local focused suite on the final source: 4 passed.
• Case-status CSV has 63 rows and three columns per row; parity-badge check and git diff --check pass.
• Full Python CI is running for the revised exact head.