Enable distributed tests for ROCm (#81751)
We're investigating why the distributed tests were running for >4.5hrs in CI when they were disabled: https://github.com/pytorch/pytorch/issues/80529
Our internal runs have been running within 4 hrs for distributed shard1+shard2.
This PR enables the distributed job to see if the runtime has improved at all.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/81751
Approved by: https://github.com/malfet