[2407.00047] One Queue Is All You Need: Resolving Head-of-Line Blocking in Large Language Model Serving