llama.cpp fixes to run Bonsai 1-bit models on CPU (incl AVX512) and AMD GPUs

1-bit quantization is essential for consumer hardware reach, but lacks cross-platform parity. Build a high-performance CPU backend kernel optimized for AVX512 specifically for the Bonsai architecture.

Suggested repo: bonsai-lite

"Run 1-bit giants on your laptop: Hyper-optimized CPU kernels for Bonsai."

Estimated effort: 120h