fix: InterpolatedStructure contact/dip selection and StructuralPoint strike wraparound

[lachlangrose]

Summary

Three correctness bugs in the two most-used thickness calculators (InterpolatedStructure is the project default; StructuralPoint is the other user-facing option):

• InterpolatedStructure picked the wrong basal contact segment. shapely.shortest_line(point, top_contact_geometry) broadcasts across the list of top-contact geometries and returns one candidate line per list entry; the code always took short_line[0]. Whenever a basal contact is split into more than one feature (common across faults or map-sheet edges), the thickness estimate silently used an arbitrary segment instead of the nearest one. Now the candidate with the minimum length is selected.
• InterpolatedStructure sampled dip from the wrong unit. The measured line spans stratigraphic_order[i + 1] (the unit whose thickness is being computed), but dip was pulled from interpolated points in stratigraphic_order[i] (the overlying unit). Dip is now sampled from the unit actually being measured.
• StructuralPoint rejected valid strike matches near due north. The strike-allowance check compared strike1/strike2 against a plain numeric range (strike - allowance, strike + allowance). Strike is a compass bearing that wraps at 360°, so e.g. a measured strike of 5° and a contact strike of 350° (15° apart) failed the check even though they're well within the 30° default tolerance. Replaced with an angular-difference comparison that handles wraparound.

Test plan

• pytest tests/thickness — all 3 existing suites (InterpolatedStructure, StructuralPoint, ThicknessCalculatorAlpha) pass unchanged
• pytest tests/project/test_thickness_calculations.py passes unchanged

[lachlangrose]

Closing — map2loop now lives in packages/map2loop of Loop3D/LoopStructural, and this repository is being archived. Ported to Loop3D/LoopStructural#327