[launch-latency] Add tvm-ffi to cutedsl

[xuzhao9]

Cutedsl now supports tvm-ffi: https://docs.nvidia.com/cutlass/latest/media/docs/pythonDSL/cute_dsl_general/compile_with_tvm_ffi.html

$ LD_LIBRARY_PATH=$HOME/.conda/envs/py312/lib python run.py --op launch_latency --only nop_python_function,nop_triton_kernel,nop_triton_compiled_kernel_run,nop_cutedsl_tvm_ffi,nop_cutedsl --force
  x_val    nop_python_function-walltime    nop_triton_kernel-walltime    nop_triton_compiled_kernel_run-walltime    nop_cutedsl_tvm_ffi-walltime    nop_cutedsl-walltime
-------  ------------------------------  ----------------------------  -----------------------------------------  ------------------------------  ----------------------
      0                     2.58363e-05                    0.00630748                                 0.00326835                     0.00654917               0.00668514
     19                     2.85705e-05                    0.0138997                                  0.00484651                     0.000710644              0.0444969
average                     2.72034e-05                    0.0101036                                  0.00405743                     0.00362991               0.025591

  x_val                                                                                                                                                      nop_cutedsl-walltime_kineto_trace                                                                                                                                              nop_cutedsl_tvm_ffi-walltime_kineto_trace
-------  -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------  -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
      0  https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/tritonbench/devgpu089.cco2.facebook.com_2596165.1764990930051555915.pt.trace.json.gz&bucket=pyper_traces  https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/tritonbench/devgpu089.cco2.facebook.com_2596165.1764990932868898706.pt.trace.json.gz&bucket=pyper_traces
     19  https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/tritonbench/devgpu089.cco2.facebook.com_2596165.1764990935475979650.pt.trace.json.gz&bucket=pyper_traces  https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/tritonbench/devgpu089.cco2.facebook.com_2596165.1764990939806473026.pt.trace.json.gz&bucket=pyper_traces
average

Triton:

  x_val                                                                                                                                                nop_triton_kernel-walltime_kineto_trace                                                                                                                                   nop_triton_compiled_kernel_run-walltime_kineto_trace
-------  -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------  -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
      0  https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/tritonbench/devgpu089.cco2.facebook.com_2945792.1764991259550770338.pt.trace.json.gz&bucket=pyper_traces  https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/tritonbench/devgpu089.cco2.facebook.com_2945792.1764991262721098781.pt.trace.json.gz&bucket=pyper_traces
     19  https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/tritonbench/devgpu089.cco2.facebook.com_2945792.1764991265462897437.pt.trace.json.gz&bucket=pyper_traces  https://www.internalfb.com/intern/perfdoctor/trace_view?filepath=tree/traces/tritonbench/devgpu089.cco2.facebook.com_2945792.1764991270028992898.pt.trace.json.gz&bucket=pyper_traces
average