add support for llama.cpp local server

[nischalj10]

llama.cpp is much faster than ollama. it also provides an open ai api compatible local server. this should be a much better way to package local models in desktop apps and would be a great addition to the repo.

[nischalj10]

update - llama.cpp has stopped supporting VLMs on its server. thus, not integrating it currently ggml-org/llama.cpp#5882