update fsdp1/2 state dict - save replicated items on all shards

[zyaoj]

What does this PR do? Please describe:

When using fsdp1/2, replicated items (e.g. model buffer) are only saved to rank 0 (fsdp1, fsdp2). However, when continue training from sharded model checkpoints from fsdp1/2, we don't retrieve and load those replicated items from sdp_00.pt. This PR fixes the subtle bug by retiring the small optimization of model buffer de-dup.

Does your PR introduce any breaking changes? If yes, please list them:
N/A

Check list:

• Was the content of this PR discussed and approved via a GitHub issue? (no need for typos or documentation improvements)
• Did you read the contributor guideline?
• Did you make sure that your PR does only one thing instead of bundling different changes together?
• Did you make sure to update the documentation with your changes? (if necessary)
• Did you write any new necessary tests?
• Did you verify new and existing tests pass locally with your changes?
• Did you update the CHANGELOG? (no need for typos, documentation, or minor internal changes)