Compare mixed and Galerkin testing against exact polarization/Fresnel interfaces and independent driven FDFD solves. Show complete-basis recovery, omitted-channel power, and the limits of longitudinal refinement, orientation averaging, and Euclidean least squares. Keep the default formulation unchanged and document its approximation limits.
Validation: 1689 tests passed, 584 skipped; Ruff and documentation build passed.
Require a finite negative ln_R_per_layer before coefficient evaluation. Document None entries as the supported way to disable a face, and cover zero, signed zero, positive, and nonfinite targets with and without frequency shift.
Validation: 209 focused tests passed; Ruff passed.
Delegate uniform-grid PML faces to stretch_with_scpml, using physical-thickness scaling and electric-center/magnetic-edge samples. Cover disabled axes, touching faces, real profiles, and final stretch caps.
Validation: 200 focused tests passed; Ruff passed.
