Benchmark Dataset and Analysis Scripts: SAH versus Morton-Code BVH Construction for Real-Time Broadphase Collision Detection (JOIV Manuscript #5338, Revision 2)

Supplementary data package for the second-round revision of JOIV manuscript #5338, "SAH versus Morton-Code BVH Construction for Real-Time Broadphase Collision Detection: A Controlled CPU Benchmark". The package contains the extended benchmark campaign CSVs covering three BVH construction engines (DynamicBVH, AdaptiveSAH, LBVH) on three high-resolution Stanford meshes (Bunny 69,451 triangles; Dragon 871,414 triangles; Happy Buddha 1,087,716 triangles) at instance counts from 20 to 1000 with 19 samples per cell, broadphase correctness validation outputs (Jaccard index against a brute-force reference, 450/450 rows at 1.0), construction-stage timing breakdown CSVs, AABB aspect-ratio measurements, and the numpy-only analysis scripts that generate every statistic and figure in the revision (bootstrap 95% confidence intervals with 10,000 resamples at seed 1337, permutation omnibus tests, scaling fits). Source repository: https://github.com/RazordousJR/BVH_CollisionDetection. Exact source commits are recorded in the README inside the package.