Fix BC by adding aten::_native_multi_head_self_attention (#72429)
Summary:
Forward fixes https://hud2.pytorch.org/minihud?name_filter=linux-xenial-py3.7-gcc5.4%20/%20test%20(backwards_compat,%201,%201,%20linux.2xlarge)
```
The PR is introducing backward incompatible changes to the operator library. Please contact PyTorch team to confirm whether this change is wanted or not.
Broken ops: [
aten::_native_multi_head_self_attention(Tensor query, Tensor qkv_weight, Tensor qkv_bias, Tensor proj_weight, Tensor proj_bias, Tensor? mask=None) -> (Tensor)
]
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/72429
Reviewed By: albanD
Differential Revision: D34043480
Pulled By: janeyx99
fbshipit-source-id: 7db8c682c7d5c3bd911a87d21670b5bd2f3ad5a1
(cherry picked from commit 0985ebb7f1ac2fc7e6a1e0f387b7771dbe19d31f)
