SIMD-optimized hypervector operations.
Provides AVX2, x86-SSE, and ARM NEON paths for common HDC primitives.