Update CMakeLists.txt

fix: fix bug of t5 beam search (NVIDIA#410)

1. fix bugs of cum_log_probs and output_log_probs of t5 beam search
2. fix bug of beam search for large beam width
3. fix bug of misaligment of beam search penalty kernel

fix: fix fmha kernel assert bug

fix: add cutlass submodule

Update stop_criteria_kernels.cu

fix: fix bug of length_criterion

Update stop_criteria_kernels.cu

fix: fix bug of length_criterion

Update T5TritonModel.cc

fix bug of T5TritonModel

feat: update v5.1 (NVIDIA#281)

feat: update v5.0

Merge branch 'dev/v5.0_beta_update' into dev/v5.0_beta