Add support for llamacpp min_p sampler

[JoseConseco]

https://github.com/ggerganov/llama.cpp/pull/3841

It supposed to give better results compared to top_k, top_p. I tried to add this min_p - parameter to llama options, but it was unrecognized.

[fast-90]

I would love to see this in Ollama.

[Robitx]

PR: https://github.com/jmorganca/ollama/pull/1825

[FellowTraveler]

I think the PR needs to be rebased again because it says "can't be automatically merged." Unless it's waiting on something else in the workflow?