[codex] Make rumple a gridded metric

[iosefa]

Summary

Make calculate_rumple consistent with the other gridded forest-structure metrics.

The released v0.4.1 implementation calculated one scalar rumple value from an existing CHM. This changes the public API so callers provide point-cloud data and voxel_resolution, and receive a 2D rumple raster plus extent, matching the pattern used by CHM/PAI/FHD-style outputs.

Changes

• Change calculate_rumple(points, voxel_resolution, min_height=None) to return (rumple, extent).
• Calculate rumple independently per output grid cell from local point TINs using X, Y, and HeightAboveGround.
• Add metric="rumple" support to process_with_tiles.
• Update rumple docs to describe the gridded point-cloud workflow.
• Replace scalar-CHM rumple tests with gridded point-cloud tests.

Ground Area

Within each output cell, ground area is now the summed projected planar area of valid Delaunay triangles formed by the points in that cell. It is not the full rectangular cell footprint. This keeps a flat but irregularly sampled point surface at rumple 1.0, while cells with fewer than three unique point locations return NaN.

Validation

• python -m pytest tests/test_calculate.py -q passed: 54 passed, 1 warning.
• python -m py_compile pyforestscan/calculate.py pyforestscan/process.py pyforestscan/__init__.py
• git diff --check