Refactor operator dispatch framework across different Tensors.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/77707
Refactor to clean up the following pieces:
1) Consolidate decorators to use a common way to look up operator tables.
2) Move a bunch of utilities to `op_registry_utils` and `common_op_utils` and
reuse them across ShardedTensor, ReplicatedTensor and PartialTensor.
Differential Revision: [D36465639](https://our.internmc.facebook.com/intern/diff/D36465639/)
Approved by: https://github.com/wanchaol, https://github.com/fduwjj
