Add 13B and 4_0 model for metal

[siilats]

use

pip uninstall llama-cpp-python -y
CMAKE_ARGS="-DLLAMA_METAL=on" pip install -U llama-cpp-python --no-cache-dir
pip install 'llama-cpp-python[server]'

to install metal version. only works on 4_0 quant models