[jacksteamdev]

Summary

Actual native + rebuilt Wasm chart APIs
  -> pinned independent SuperNOVAS reference
  -> numerical rows + target outcomes + retained failures

Adds opt-in QC for fifteen planetary/derived bodies, including Ceres/Chiron and osculating nodes with symbolic South declination, reported offsets, original IAU 2000B nutation, twelve cusps and four angles. It validates reference provenance/models/echoed inputs and records source/native/Wasm hashes, all cases and reference geometry. Native/Wasm attempts are independent.

Based on fix/lahiri-full-nutation-import at 9f84a6e9ec815fb52418f26d98a6ccdd78d688dd (PR #15). Library calculations and regression fixtures are unchanged. Effective TT/UT1 diagnostics remain distinct from planetary checks at physical TT.

Evidence

• Before: no independent native/Wasm benchmark gate.
  After: 831 cases, including 423 chart inputs, complete 53,457 numerical checks with zero numerical failures. This includes 4,932 Ceres/Chiron checks against pinned Horizons inputs.
• The gate exits 1: twelve polar inputs are rejected by each target as overlapping houses (24 errors); each target completes 411 charts. Inputs and independent ordered Porphyry fallback geometry are retained. See tools/verification/novas-qc-findings.md.
• Dataset SHA-256: 06d4c26e589207bf479dbd217fafb3e4d6c559626425b156912e39a4d859487b. Tampered node models and geometry identity fail with exit 2; numerical perturbations fail with exit 1.

Merge Danger

Door: two-way

Blast Radius: tooling

Coverage remains partial: mean-node/geometric-South modes and True Sky are excluded. Regular PR CI is distinct from the failing opt-in full-data QC gate. The new manual hosted QC workflow has not been dispatched.