[cmccomb]

Summary

• Replace the per-DOF and per-heading linear-frequency interpolation loops with batched SciPy degree-one interpolation of the real and imaginary coefficient tensors.
• Preserve the in-range frequency check, (6, headings) output, rho * g scaling, cubic-spline path, and periodic heading interpolation.

Verification

• Compared the previous and new implementations on the pinned OSWEC HDF5 (500 frequencies, 36 headings) at both endpoints and five interior frequencies. Maximum absolute differences were 2.33e-10 for real coefficients and 4.66e-10 for imaginary coefficients and evaluated forces; shapes matched. Out-of-range frequencies still reject.
• python -m pytest -q tests/test_oswec_passive_yaw_portable.py tests/test_oswec_passive_yaw_parity.py tests/test_oswec_regular_yaw_portable.py tests/test_variable_yaw_heading_bank.py tests/test_passive_yaw_configuration.py: 15 passed, 11 skipped locally (live MATLAB output variables unavailable). The portable tests use checked MATLAB fixtures.
• Targeted live MATLAB OSWEC passive-yaw pairing on this commit: both paired cases passed against freshly generated pinned-source output.
• git diff --check passed.

Closes #631