MLX engine support

[onderceylan]

Hi there! First of all, amazing project - I'm a big fan! Thanks for the hard work.

About the request, I was blown away with the local inference speed via Apple Silicon (M1,M2,M3) on recent LM Studio release using https://github.com/lmstudio-ai/mlx-engine.

Is this something we can bring to the MLC AI community and the WebLLM project?