GitLab

This one log line was triggering a single line llama.log to be generated
in the pwd of the server

This changes the model for llama.cpp inclusion so we're not applying a patch,
but instead have the C++ code directly in the ollama tree, which should make it
easier to refine and update over time.