[PyTorch] Support scaled + clamped SwiGLU in te.ops and enable fused MXFP8 grouped MLP#2855

ksivaman merged 7 commits intoNVIDIA:mainNVIDIA/TransformerEngine:mainfrom

ksivaman:fused_mxfp8_mlp_gegluksivaman/TransformerEngine-1:fused_mxfp8_mlp_gegluCopy head branch name to clipboard