Add high level llama-cpp-python logits processor

[dtiarks]

Implement an outlines based logits processor for llama-cpp-python according to the interface in docs: https://llama-cpp-python.readthedocs.io/en/latest/api-reference/#llama_cpp.LogitsProcessor

[rlouf]

This would be a much better way to integrate with llama.cpp since it would require less maintenance on our end.

[dtiarks]

Yeah, I guess this is also true for transformers and others (👀 TensorRT-LLM).