[Triton-MLIR][BACKEND] Tiny patch for MMAv1 and code clean (#964)

This PR:

- Several fix on MMAV1 code
- Remove the env `TRITON_STATIC_LOOP_UNROLLING` in v100 CI since the
pipeline pass works now
- some code clean

.github/workflows/integration-tests.yml

@@ -88,9 +88,7 @@ jobs:
       - name: Run python tests on V100
         if: ${{matrix.runner[0] == 'self-hosted' && matrix.runner[1] == 'V100'}}
         run: |
-          # TODO[Superjomn]: Remove the forloop-unroll setting after pipeline pass works
           cd python/tests
-          export TRITON_STATIC_LOOP_UNROLLING=1
           pytest test_gemm.py::test_gemm_for_mmav1
 
       - name: Run CXX unittests