lalsimutils: phi12, chi_p_vec, vectorized in-plane spin coordinates (O4d port of #203)

[oshaughnessy-junior]

rift_O4d port of oshaughn#203.

extract_param gains phi12 (listed in valid_params, never implemented) and chi_p_vec, the vector sum of the in-plane terms in chi_p. The spherical-spin branch of convert_waveform_coordinates builds chi1_perp, chi2_perp, phi12, SOverM2_perp, DeltaOverM2_perp and chi_p_vec vectorized, in the L frame only.

Review fixes, also applied to #203: the block accepts CIP's object arrays (the default sampler crashed), keeps the enforce_kerr rule, and returns phi12 = 0 at zero in-plane spin. chi_p_vec is not in valid_params, since grid readers assign every listed column.

No defaults change. The CIP width results (RIFT_roboto_paper analyses/transverse_convergence/RESULTS_ring_coordinate_2026-09-30.md) were measured on rc3, not O4d.

Tests (ldas-grid, IGWN python 3.11, numpy):

• test_ring_coordinates.py: 7 passed
• test-core-units.sh: PASS, 608 collected, 595 passed, 0 failed; floors 605/592
• roster and vector-coordinate tests pass; existing coordinate sets bit-identical to base
• CIP with the default sampler and the new coordinates: exit 0

🤖 Generated with Claude Code

[oshaughnessy-junior]

[P2] Preserve enforce_kerr when moving ring coordinates out of the fallback

At head 0f933e7a7a2a1f772eb4feba21e03303d29a80e1, removing these outputs from coord_names_reduced lets the existing early return bypass the only enforce_kerr check, which runs in the per-row fallback. This changes the behavior of existing SOverM2_perp/DeltaOverM2_perp outputs and can admit super-Kerr rows through CIP's --downselect-enforce-kerr coordinate transformation. Reproduced with L convention, low_level_coord_names=['mc','delta_mc','chi1','cos_theta1','phi1','chi2','cos_theta2','phi2'], x_in=np.array([[20.,.3,1.2,.2,1.,.5,.4,2.]]), coord_names=['SOverM2_perp'], and enforce_kerr=True: parent f0d90b21 returns [[-inf]], whereas PR head returns [[0.52919969]]. Adding chi_p to the requested head outputs returns [[-inf,-inf]] because it forces fallback, so row validity now depends on the output feature list. Apply the Kerr rejection mask before the vectorized early return (or retain fallback when enforcement is requested), and cover both super-Kerr and valid rows in a regression test.

[oshaughnessy-junior]

[P2] Use one phi12 convention at zero transverse spin in both conversion paths

Adversarial review of head 0f933e7a7a2a1f772eb4feba21e03303d29a80e1: the new vector path at lalsimutils.py:6161 subtracts the input azimuths even when a transverse spin is zero, whereas extract_param('phi12') at line 1396 derives its azimuth from the Cartesian spin. These give different finite fitting coordinates for the same physical configuration.

Concrete reproduction (NumPy backend):

LOW = ['mc','delta_mc','chi1','cos_theta1','phi1','chi2','cos_theta2','phi2']
x = np.array([[20., .3, 0., .2, 1., .5, .4, 2.]])
convert_waveform_coordinates(x, coord_names=['phi12'], low_level_coord_names=LOW)
# [[1.]]
# Build the equivalent Cartesian ChooseWaveformParams:
# spin 1 = (0,0,0), spin 2 azimuth = 2 radians
P.extract_param('phi12')
# 2.0

The same mismatch occurs with nonzero chi1 and cos_theta1=+1 or -1. All other ring coordinates agree for these cases. phi12 is physically undefined at a vanishing transverse spin, so a canonical policy must be selected; this is not an argument for a particular angle. The issue is that CIP assembles training fit coordinates using P.extract_param (util_ConstructIntrinsicPosterior_GenericCoordinates.py:2391) and obtains sampling fit coordinates through the vector converter. Exact aligned/zero-spin likelihood grid points can therefore be encoded differently from the sampling path, even with identical physical spins.

Please apply a common canonical policy in both paths and add zero-spin/aligned-pole cases to the agreement test. The two submitted tests pass, as do 1,000 additional random nondegenerate draws and individual-coordinate subsets; their continuous random draws never exercise these boundaries.

[interhost-cross-review-canary-v2]

Independent automated review completed at the recorded exact commit. Detailed findings were withheld from public output by the private-context egress policy and require private human declassification.

[interhost-cross-review-canary-v2]

Independent automated review completed at the recorded exact commit. Detailed findings were withheld from public output by the private-context egress policy and require private human declassification.

[interhost-cross-review-canary-v2]

Independent automated review completed at the recorded exact commit. Detailed findings were withheld from public output by the private-context egress policy and require private human declassification.