Performance model update for LLMServingSim
New features
- Support GPU with a performance model
- Auto config generator (network and memory)
- Verbose option for more detailed log
- More metrics (queuing_delay, TTFT, TPOT)
- Refactored code structure for readability