llama.cpp

server: run sampling in a threadpool

#24914

Open