* small updates to llm performance tests regarding model handling * Small fix
* add xpu latency job * update install way * remove duplicated workflow * add perf upload