Apple Silicon and macOS VMs: Faster LLM Inference with llama.cpp