[https://nvbugs/5537996][fix] Let KV cache manager block initialization be aware whether it is doing a dry run or not#9093

eopXD merged 2 commits intoNVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

eopXD:fix-max_tokens-settingseopXD/TensorRT-LLM:fix-max_tokens-settingsCopy head branch name to clipboard