Avoid using block_m=64 when having mxfp8 downcast in epilogue

[qnie-oai]

New contributor declaration

• I am not making a trivial change, such as fixing a typo in a comment.

• I have written a PR description following these
  rules.

• I have run pre-commit run --from-ref origin/main --to-ref HEAD.

• Select one of the following.

  • I have added tests.
    • /test for lit tests
    • /unittest for C++ tests
    • /python/test for end-to-end tests
  • This PR does not need a test because FILL THIS IN.

• Select one of the following.

  • I have not added any lit tests.
  • The lit tests I have added follow these best practices,
    including the "tests should be minimal" section. (Usually running Python code
    and using the instructions it generates is not minimal.)

pytest ./python/triton_kernels/tests/test_matmul.py -rs -vv -k "True-False-True-False-None-128-720-576-768-ragged-mxfloat8_e4m3fn-mxfloat4_e2m1-10-1-True-None-False-False-False-True-swiglu_opt"

• failed with numerics gap with ref

pytest ./python/triton_kernels/tests/test_matmul.py -rs -vv -k "True-False-True-False-None-128-768-512-1024-ragged-mxfloat8_e4m3fn-mxfloat4_e2m1-10-1-True-None-False-False-False-True-swiglu_opts"

• passed test

[lezcano]

Making the block larger should increase the shmem usage, so this change on its own does not make sense to me.
Perhaps the right change is to change the shmem computation we do when setting num_stages?