chore(tier3): refreeze baselines snapshot (track 19 untracked tests + LSODE)

[simnaut]

Summary

Refreezes the Tier 3 baselines.json regression snapshot from a clean full
suite run, closing a drift gap between the frozen baseline (78 entries) and
the actual set of passing tier3 tests (99).

baselines.json is a manual tier3_report --freeze-baselines artifact —
CI never refreshes it. The tier3_baseline_diff gate only fails on
regressions of entries already present; brand-new tier3 tests are reported
as informational new and pass. So every tier3 test added since the last
freeze accumulated unguarded against future regression.

Regenerated via:

rm -rf target/tier3_crossval
cargo nextest run --workspace -E 'test(tier3_)'   # 233 passed
cargo run -p astrodyn_verif_jeod --bin tier3_report -- --freeze-baselines

What changes (78 → 99 entries)

+19 tracked-but-unfrozen tests now gain regression coverage:

csr_compare_gravity_octants, earth_moon_rosetta, mars_orb_init_phobos,
mars_phobos, nesc_cc8_nrho, sim_attach_detach_trajectory_simple,
sim_complex_attach_detach_pre_attach_trajectory,
sim_compute_child_derivative_{full,pre_attach}_trajectory,
sim_dyncomp_run_attach_to_ref_frame, sim_kinematic_propagation_simple,
simulation_relative_{a_rot_no_trans,ab_rot_ab_trans,no_rot_ab_trans},
simulation_run{2_lvlh_rot_init_propagation,6c_plane_change,6d_departure,9b_torque_initial_rate,10b_gravity_torque_circular_rate}.

1 existing entry tightened — tier3_simulation_lsode_default:

component	old ceiling	new
position x/y/z	9.5e3 / 9.1e3 / 6.0e3 m	9.2e-4 / 8.7e-4 / 5.8e-4 m
velocity x/y/z	1.1e1 / 9.9e0 / 6.6e0 m/s	1.0e-6 / 9.6e-7 / 6.4e-7 m/s

The old ceiling was frozen while LSODE was a stub; #616/#617 implemented the
stiff BDF integrator, so the real error is now sub-millimeter. A ~7-order
accuracy regression could previously pass unnoticed.

Safety

• All deltas are tightenings (informational) or new entries — no
  baseline was loosened.
• scripts/check_baseline_widening.sh vs origin/main:
  0 error(s), 0 warning(s), 6 tightening(s), 21 new, 0 removed.
• tier3_baseline_diff: OK (99 matched; 0 allowed-missing; 0 new).
• tier3_sim_attach_mass is unchanged from test(tier3): add 9 SIM_verif_attach_mass RUNs (#99 Bucket B) #624 (byte-identical).

Follow-up worth considering (not in this PR)

There's no forcing function keeping the frozen set in sync with the tier3 test
set. A CI check that fails when a tier3_ test exists without a baseline entry
(rather than reporting new as a passing notice) would prevent recurrence.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

Updates the Tier 3 JEOD cross-validation baseline snapshot to reflect the current set of passing tier3_ tests, so newly-added Tier 3 tests gain regression coverage and LSODE’s baseline reflects the now-real solver accuracy.

Changes:

• Refreezes Tier 3 baselines from 78 → 97 tracked tests (adding the previously “new/unfrozen” Tier 3 tests).
• Tightens the tier3_simulation_lsode_default baseline thresholds to sub-millimeter / ~1e-6 m/s levels.
• Regenerates the human-readable baselines.md to match the updated JSON snapshot.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
crates/astrodyn_verif_jeod/test_data/baselines.json	Adds 19 new baseline entries and tightens LSODE default tolerances for Tier 3 regression gating.
crates/astrodyn_verif_jeod/test_data/baselines.md	Regenerates the readable baseline report (count + new sections) to mirror the updated JSON snapshot.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.