llama.cpp

ci: bench: support sse and fix prompt processing time / server: add tokens usage in stream OAI response

#6495

Merged