[state_dict][2/N] Let distributed.state_dict accepts single optimizer (#111107)
It's quite annoying that users have to create a tuple of optimizers even if there is only one optimizer. This PR makes most users' life easier.
Differential Revision: [D50209704](https://our.internmc.facebook.com/intern/diff/D50209704/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/111107
Approved by: https://github.com/wz337
ghstack dependencies: #111106
