[FSDP] Add HSDP parity unit test (#106131)
With >= 4 GPUs:
```
python -m pytest test/distributed/fsdp/test_fsdp_hybrid_shard.py -k test_fsdp_hybrid_shard_parity
```
Differential Revision: [D47852458](https://our.internmc.facebook.com/intern/diff/D47852458)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/106131
Approved by: https://github.com/rohan-varma