Add VMEC->Boozer converter (boozer_sub) to libneo, pinned against SIMPLE

[krystophny]

Summary

Bring the VMEC-to-Boozer spline converter into libneo as the canonical module boozer_sub. The converter currently lives duplicated and diverged in SIMPLE (src/boozer_converter.F90) and rabe (src/vmec/boozer_converter.F90); their numeric cores are identical. This lands one libneo copy so both downstreams can drop their forks (cutover PRs follow).

This is the first step of the hub-and-spoke converter consolidation (see #305): one in-memory Boozer representation that every format reads into and writes out of.

Scope

• New src/boozer_converter.F90 (module boozer_sub), added to the neo library. It depends only on modules libneo already provides: spline_vmec_sub (incl. vmec_field, splint_vmec_data, metric_tensor_vmec), boozer_coordinates_mod, vector_potentail_mod, new_vmec_stuff_mod, interpolate, plag_coeff_sub, binsrc_sub, spl_three_to_five_sub, chamb_mod, diag_mod.
• Public API: get_boozer_coordinates(vmec_file, radial_spline_order, angular_spline_order, grid_refinment), splint_boozer_coord, reset_boozer_batch_splines. Carries the contravariant sqrt(g^ss) metric and the optional B_r reconstruction (use_B_r).
• New regression test test_boozer_converter_vs_simple pinning the converter output against SIMPLE.

SIMPLE-only supersets (the vmec_to_boozer/boozer_to_vmec angle transforms, the OpenACC boozer_state device path, and the chartmap load/export spoke) are intentionally not in this PR; they land as additive follow-ups before the SIMPLE cutover. Behavior is unchanged for existing libneo code: nothing else in libneo calls boozer_sub yet.

Verification

Before: libneo had no VMEC->Boozer converter, so the pin could not be built.

After, the converter reproduces SIMPLE's published reference values at five (s, theta, phi) points on the LandremanPaul2021 QA reference wout (|B|, sqrt(g), bder, hcovar, hctrvr, hcurl; reltol 1e-6, abstol 1e-11):

$ ctest -R "setup_vmec_wout|test_boozer_converter_vs_simple" --output-on-failure
    Start 30: setup_vmec_wout
1/2 Test #30: setup_vmec_wout ...................   Passed    0.24 sec
    Start 48: test_boozer_converter_vs_simple
2/2 Test #48: test_boozer_converter_vs_simple ...   Passed    1.56 sec

100% tests passed, 0 tests failed out of 2