llama.cpp

server: add KV cache metrics

#24010

Open