Stale rotated free-slip datum guard tests in test_0641 (nonzero datum is now implemented, #458)

[lmoresi]

tests/test_0641_wave_c_api_shims.py::TestRotatedFreeslipValueFirst still asserts that a non-zero rotated free-slip datum raises NotImplementedError:

• test_nonzero_datum_not_implemented
• test_symbolic_possibly_nonzero_datum_not_implemented

#458 (merged 2026-07-29) implemented the prescribed-normal datum, so add_rotated_freeslip_bc(1.0, "Top") is now accepted and both tests fail on development.

FAILED tests/test_0641_wave_c_api_shims.py::TestRotatedFreeslipValueFirst::test_nonzero_datum_not_implemented
FAILED tests/test_0641_wave_c_api_shims.py::TestRotatedFreeslipValueFirst::test_symbolic_possibly_nonzero_datum_not_implemented
2 failed, 82 passed

They should assert the new contract instead: a scalar/symbolic datum is accepted and recorded on _rotated_freeslip_bcs, and whatever is still genuinely rejected (vector/matrix values) still raises.

Found while validating the FMG option-bundle unification (#468 / #467); unrelated to that change and not fixed there to keep the diff scoped.

Underworld development team with AI support from Claude Code

[lmoresi]

Fixed by #487, merged as 5c904836. level_1 and tier_a is now 530 passed, 0 failed — these two were the only red tests on development.

Replaced with the contract that actually holds, read off live behaviour: a numeric datum is recorded, a symbolic one is kept (is_zero is True only for a provable zero), a provable zero records no datum, and a vector value is still a TypeError. The last two are additions rather than replacements, so "a datum is accepted now" cannot be over-read as "anything is accepted".