Add Metal backend; vendor whole llama.cpp trees

GGUF kernels: dequantize + fused gemv over packed blocks

GGUF kernels: dequantize + fused gemv over packed blocks, 9 CUDA variants