Update chunk_dtensor to support HYBRID_SHARD (#119481)
Fixes https://github.com/pytorch/pytorch/issues/118639.
Adds support to replicate across HSDP dimensions instead of sharding for shard placement
Pull Request resolved: https://github.com/pytorch/pytorch/pull/119481
Approved by: https://github.com/Skylion007, https://github.com/wz337
