MPS Inference

Transformers Adds Fast GGUF Model Inference on Apple Silicon

Transformers adds efficient GGUF inference on Apple Silicon, using ggml kernels to run quantized Qwen3.5 and Qwen3.8 models through familiar APIs.