Configure ollama for local model serving

This skill is for DevOps & Infrastructure and helps you deploy gemma 4 locally on mac, set up llama.cpp with gguf quantization, and expose openai-compatible api locally.