Support MLA nvfp4 quant for Deepseek for max perf#582

cjluo-nv merged 1 commit intoNVIDIA:mainNVIDIA/Model-Optimizer:mainfrom

binghanc:binghanc/dpskr1_nvfp4_v3binghanc/TensorRT-Model-Optimizer:binghanc/dpskr1_nvfp4_v3Copy head branch name to clipboard