Implemented MeshCache — caches AobaAPI topology traversals (verts/edges/faces/loops) behind a dirty flag, rebuilt only on topology change. Eliminates ~2F+7 heap allocations per frame (where F = face count). Persistent wireframe sort buffer avoids EdgeDraw churn. Addresses ARCH-REVIEW Issue #2 (per-frame allocation bottleneck for Pi target).