fast_winding_number: point-cloud and pre-built tree support (#292)

[alecjacobson]

Closes #292.

The binding only exposed the (V, F, Q) triangle-soup one-shot. This adds the
rest of libigl's fast_winding_number API so point clouds and reusable
acceleration structures are supported (enabling full signed-distance workflows
and multiple query sets against one build).

New API

Point-cloud one-shot

W = igl.fast_winding_number(P, N, A, Q, expansion_order=2, beta=2.0)

Reusable triangle-soup BVH — build once, query many times:

bvh = igl.FastWindingNumberBVH()
bvh.init(V, F, order=2)
W1 = bvh.winding_number(Q1)
W2 = bvh.winding_number(Q2)   # no rebuild

Point-cloud octree precompute — reuse the Taylor expansion across query sets
(builds on the already-bound igl.octree):

point_indices, CH, CN, _ = igl.octree(P)
CM, R, EC = igl.fast_winding_number_precompute(P, N, A, point_indices, CH, 2)
W = igl.fast_winding_number(P, N, A, point_indices, CH, CM, R, EC, Q, beta=2.0)

The original igl.fast_winding_number(V, F, Q) is unchanged.

Testing

test_fast_winding_number exercises all four paths on a unit-sphere point
set/mesh: interior queries → ~1, exterior → ~0, and the BVH and cached-octree
results match their one-shot equivalents. Full suite: 92 passed locally.

🤖 Generated with Claude Code

[alecjacobson]

@copilot resolve the merge conflicts in this pull request

[Copilot]

@copilot resolve the merge conflicts in this pull request

Done — merged origin/main into the branch and resolved the conflict in tests/test_all.py by keeping both test_fast_winding_number (from this PR) and test_vertex_components_from_adjacency_matrix (from main). Commit: 1c1085f.

[alecjacobson]

@copilot resolve the merge conflicts in this pull request

[Copilot]

@copilot resolve the merge conflicts in this pull request

Resolved — merged the latest origin/main and kept both test_fast_winding_number (from this PR) and test_ambient_occlusion (new in main). Commit: 413cc37.