trunk/bf9ea7b0a83587fa2b027958323f1c1bf9ced16e: [Inductor][NVGEMM] Enable Epilogue Fusions (#186183)

Adds epilogue fusion (relu, sigmoid, tanh, exp, add, sub, mul, div,

dtype cast) to the NVGEMM backend via cutlass_api EpilogueArguments.

The scheduler's fusion loop recognizes NVUniversalGemmCaller, benchmarks

fused vs unfused EFC kernels, and selects the winner.

MultiTemplateBuffer gains swap/finalize_as_nvgemm_caller for the

benchmark swap. When NVGEMM can't fuse, MTBs fall through to Triton

scheduling so Triton choices can still attempt epilogue fusion.

Authored with the help of an AI assistant (Claude).

Test Plan:

python test/inductor/test_nv_universal_gemm.py -v -k Epilogue

python test/inductor/test_max_autotune.py -v -k epilogue_fusion

Pull Request resolved: #186183

Approved by: https://github.com/drisspg