[Gluon] Remove tcgen05_mma wrapper from attention tutorial (triton-lang#7536)

This was a workaround for an issue that has since been fixed.

Author: peterbell10

python/tutorials/gluon/01-attention-forward.py

@@ -16,7 +16,7 @@
     tensor_memory_descriptor,
     tma,
     mbarrier,
-    tcgen05_mma as _tcgen05_mma_impl,
+    tcgen05_mma,
     tcgen05_commit,
 )
 
@@ -173,11 +173,6 @@ def issue_async_tma_load(smem, bar, desc, offset):
     tma.async_copy_global_to_shared(desc, [offset, 0], bar, smem)
 
 
-@gluon.jit
-def tcgen05_mma(a, b, d, use_acc, mbarriers):
-    _tcgen05_mma_impl(a, b, d, use_acc=use_acc, mbarriers=mbarriers, mbarrier_preds=[True] * len(mbarriers))
-
-
 # ===-----------------------------------------------------------------------===#
 # Gluon Attention
 # ===-----------------------------------------------------------------------===#