Verify against NASA NESC Lunar Check Cases

[simnaut]

Goal

Add a new verification track that propagates the NASA NESC Lunar Check Cases through Simulation::step() and compares against the published reference trajectories.

The NESC GN&C check cases (NESC-RP-12-00767 / "Independent Orbit Propagation Verification") are a canonical, vendor-neutral benchmark for high-fidelity propagators: cislunar transfer, lunar orbit, and Earth–Moon–Sun three-body trajectories with documented ICs, force models, integrator settings, and reference ephemerides at fixed checkpoints. Passing them would give us a cross-check that is independent of JEOD/Trick and exercises lunar-regime physics our existing Tier 3 fleet only touches lightly.

Scope

• Locate and ingest the NESC reference data set (ICs + checkpoint state vectors) and commit a parsed mirror under crates/astrodyn_verif_nesc/test_data/ (new crate, parallel to astrodyn_verif_jeod).
• Stand up a regen binary that reads the original NESC artifacts (when $NESC_HOME or a --nesc-home path is supplied) and writes the committed fixtures, mirroring the extract_* pattern.
• Add a tier3_nesc_* test per check case driving astrodyn_runner::Simulation end-to-end from the NESC ICs and asserting position/velocity (and attitude, if the case specifies it) against the checkpoint vectors with tolerances set per the No Half-Baked Implementations rule (match the case's force model exactly — point mass vs spherical harmonics, third-body set, integrator, step size).
• Wire a corresponding parity-trait sibling so the Bevy adapter inherits coverage transitively (per the issue Make bevy_parity tests a superset of every runner↔JEOD Tier 3 scenario #389 infrastructure), or add a KNOWN_PARITY_GAPS exemption with a written reason if a case can't ride the recipe path.
• Document the regen workflow in the new crate's README.md and link it from CLAUDE.md alongside the JEOD regen instructions.

Out of scope

• Adding NESC fixtures to the existing astrodyn_verif_jeod crate. Keep the NESC track in its own crate so the JEOD vs NESC provenance never gets mixed.
• Any new physics. If a check case exposes missing physics, that's a separate issue and a port — not an approximation in the test.

Open questions

• Which release of the NESC check cases is canonical? (There are several revisions floating around; we should pin one.)
• Do the published cases assume DE405/DE421/DE440 ephemerides? We need to match exactly what the reference used, not what we happen to have loaded.
• Is the attitude case set worth including in the first cut, or should it be a follow-up after the translational cases land?