feat(hslm): add comprehensive SIMD benchmark suite

- Patch #5 (AVX2 32-wide) performance measurement
- Benchmarks: dotProduct (fixed vs adaptive), sparse dot (zero-skip), ternary matvec (243×729)
- System info display showing SIMD capabilities (AVX2/NEON/SSE2)
- Platform-aware: skips speedup tests on non-AVX2 architectures
- All 52 tests pass (includes f16_utils, sparse_simd, simd_config)

Run: zig test src/hslm/simd_bench.zig

φ² + 1/φ² = 3 | TRINITY

Author: Antigravity Agent