AMQP interface to llama.cpp, like the HTTP Server

[remnin]

I love that llama.cpp comes with a HTTP server. But, we are currently building an AI infrastructure, and went with AMQP(rabbit) as our main LLM_prompt distribution/queuing system, so it would be great to skip the HTTP step, and just have llama.cpp feed of an queue, and also just specify a response(and error) queue. We could run separate queues for different LLM models.