[TRTLLM-9198][perf] Add torch.compile + multi-stream support for k-cache scatter and weight scaling#8988

yuxianq merged 6 commits intoNVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

chang-l:indexer_weight_compile_multistreamchang-l/TensorRT-LLM:indexer_weight_compile_multistreamCopy head branch name to clipboard