test_mla_helix: Add reduce_output=False to disable NCCL all-reduce

The MLA's o_proj layer performs an all-reduce by default, which fails
in the MPI worker context because NCCL groups aren't initialized.
Setting reduce_output=False disables this all-reduce operation.

Author: brb-nv

tests/unittest/_torch/modules/test_mla_helix.py

@@ -632,6 +632,7 @@ def _run_mla_distributed(
         dtype=scenario.dtype,
         config=config,
         enable_unit_test=True,
+        reduce_output=False,  # Disable all-reduce to avoid NCCL in MPI workers
     ).cuda()
 
     # ========================================================================