[#1593] Add analytic accuracy comparison smoke tests - #1603

• Tickets addressed: Follow-up to Add GMAT/Orekit accuracy comparison #1593, future-work item 6 (optional comparison smoke test).
• Review: By file
• Merge strategy: Merge (no squash)

Description

The accuracy-comparison pipeline currently lacks a short offline numerical smoke test. Add four cases in src/tests/test_accuracy_comparison_smoke.py that run the existing comparison entry point with native Basilisk propagation for 600 seconds and check an independent closed-form circular-orbit reference.

Temporary CSV files fill the GMAT/Orekit input slots, with manifests explicitly identifying them as synthetic analytic fixtures, not outputs from those tools. Cases cover both slots, Orekit-only selection, reporting deliberately offset states, and rejecting timestamp misalignment. Position and velocity tolerances are 1 cm and 0.1 mm/s. Existing platform CI collects the file automatically; no workflow or dependency changes are needed.

Verification

• python -m pytest src/tests/test_accuracy_comparison_smoke.py -q: 4 passed in 2.30 s.
• 28 existing reference-metadata tests passed.
• A negative control changed propagation gravity by 0.1%; the positive case correctly failed with a 1653.858 m position discrepancy.
• Existing CI-style collection discovers all four cases; applicable pre-commit, docstring, and whitespace checks passed.

Local checks used macOS arm64, Python 3.12, NumPy 2.3.5, and the official Basilisk 2.12.0 native wheel with the current source comparison scripts. The wider comparison run had 32 passes and one existing failure: test_orekit_only_selection_needs_no_gmat_data requires ExponentialAtmosphere.setPlanetPolarRadius, which is absent from that native wheel. The failure also reproduces when the existing test runs alone. It was deselected for the metadata-only run. A native build of current develop and the full project suite remain unverified, so this PR is a draft pending that validation.

Documentation

The new file documents its fixture provenance and invocation. No production comparison case data, reference results, or API documentation change. Test-only maintenance is exempt from the release-note snippet requirement.

Future work

Validate against a native build of current develop and run the full project suite before marking ready for review.