[TRTLLM-10171][fix] Correct attention handling in ModelConfig and KVCacheManager#10330

jaedeok-nvidia merged 3 commits intoNVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

jaedeok-nvidia:fix/vswa-with-adpjaedeok-nvidia/TensorRT-LLM:fix/vswa-with-adpCopy head branch name to clipboard