Expose --engine flag on the layup orbitfit CLI

[matthewholman]

This branch was developed with Claude Code, on top of an initial
implementation I had started.

Wires the engine choice end-to-end through layup orbitfit's
command-line entry point.

Why

PR #326 adds the universal Bernstein–Khushalani fitter as
engine="bk_native" on do_fit, but the layup-orbitfit CLI
entry point doesn't forward that argument — so even after #326
lands, running

layup orbitfit input.csv ADES_csv

always uses the Cartesian engine. This PR closes that gap so the
new engine is actually reachable from the command line:

layup orbitfit input.csv ADES_csv --engine bk_native

What changed

End-to-end plumbing:

CLI argparse --engine {cartesian, bk_native}, default 'cartesian'
  └─► orbitfit_cli reads cli_args.engine
       └─► orbitfit() takes engine= parameter
            └─► process_data_by_id forwards through **kwargs
                 └─► _orbitfit() receives engine and passes through
                      └─► do_fit(..., engine=...) (already engine-aware in #326)
                           └─► _run_fit() dispatches to the right LM

argparse's choices=["cartesian", "bk_native"] gives users an
early error on a typo'd engine name; the existing _run_fit
dispatch also raises ValueError for unknown engines if a Python
caller bypasses argparse.

Tests

• test_orbit_fit_cli_raises_with_unknown_engine: parallel to the
  existing unknown-iod test. A FakeCliArgs with engine="not_an_engine"
  triggers ValueError from _run_fit's dispatch.
• test_orbit_fit_cli_raises_with_unknown_iod: the existing test
  gets a FakeCliArgs.engine = "cartesian" field so it continues
  passing after orbitfit_cli starts reading that attribute.

Confirmed manually: layup orbitfit --help now shows

  --engine {cartesian,bk_native}
                        LM fitter to use after IOD: 'cartesian' (default;
                        classic barycentric-Cartesian LM) or 'bk_native'
                        (universal Bernstein-Khushalani fit with energy prior,
                        better-conditioned for distant short-arc targets and
                        at least as good elsewhere).

Dependencies

Stacks on feat/bk-everywhere (PR #326 — adds the engine parameter to
do_fit and the bk_native engine itself). This PR is purely the CLI-
plumbing follow-up. Once #326 lands, this PR collapses to its one new
commit.

Review Checklist for Source Code Changes

• Does pip install still work?
• Have you written a unit test for any new functions? — test_orbit_fit_cli_raises_with_unknown_engine
• Do all the units tests run successfully? — 43 BK + CLI validation tests pass locally
• Does Layup run successfully on a test set of input files/databases? — layup orbitfit --help shows the new flag; the existing test_orbit_fit_cli_raises_with_unknown_iod continues to pass
• Have you used black on the files you have updated?