FEAT: add gemma-2-it

[qinxuye]

Fixes #1750

This PR added gemma-2-it. Some notions:

1. When I ran gguf formats with llama-cpp-python==0.2.81, it crashed, and I can see the gemma-2 format is introduced lately in llama.cpp itself, so I removed gguf formats for now.
2. vllm implemented gemma-2 but it's not released, thus vllm is absent as well.
3. I added MLX format which worked well on my laptop. (Hence this PR replies on FEAT: support MLX engine #1765 )