Fix strike direction for nstk == 1 SRF segments (degrees used as radians)

[lispandfound]

Fixes #84

Problem

SrfFile.planes builds the along-strike offset for single-strike-point segments from a bearing returned by qcore.coordinates.great_circle_bearing_to_nztm_bearing, which is in degrees, but passed it straight to np.cos/np.sin:

# source_modelling/srf.py:615 (before)
* np.array([np.cos(strike_nztm), np.sin(strike_nztm), 0])

The bearing was therefore interpreted as radians and the resulting plane pointed in an essentially arbitrary direction.

Measured impact

Recovered plane.strike from a synthetic nstk == 1 segment, before this change:

header stk	recovered	error
0	87.16	87.2
20	50.24	30.2
45	44.70	0.06
90	97.33	7.7
200	103.60	96.1
330	343.26	13.7

After the change every bearing tested recovers the header strike to within 0.05 degrees.

Why CI never caught this

test_planes_nstk_1_ndip_gt_1 exercises exactly this branch, but with stk=45 — which turns out to be an accidental near-fixed-point of the degrees/radians confusion (44.758 rad reduces to 44.44 deg mod 2*pi, 0.3 degrees from the true value). The assertion plane.strike == approx(45, abs=1) passed on a coincidence. That blind spot is why the new test is parametrized over eight bearings rather than tightening the existing one.

Changes

• source_modelling/srf.py — wrap strike_nztm in np.radians, matching the convention already used at fsp.py:124-125.
• tests/test_srf.py — add test_planes_nstk_1_strike_recovered, parametrized over 0/20/45/90/135/200/270/330 degrees.

The new test needed a fixture whose point column is genuinely consistent with the header strike (the down-dip column generated along stk + 90). The pre-existing fixture hardcodes a southeast-dipping point column, so for strikes like 200 the header and the point geometry contradict each other and Plane reorders the corners, flipping the reported strike by 180 degrees — unrelated to this bug, but it makes the old fixture unusable for a strike sweep.

Verification

• New test fails on all 8 parameters without the fix, passes with it.
• pytest tests/test_srf.py — 24 passed.
• ruff check, ruff format --check, numpydoc lint clean.

Only line 615 was affected; it is the sole trig call in srf.py, and the ndip == 1 branch delegates to Plane.from_centroid_strike_dip, which converts internally.

🤖 Generated with Claude Code