Let TorchCollective works on the torch.distributed WORLD process group by default

[belerico]

What does this PR do?

With this PR TorchCollective works by default on the torch.distributed.group.WORLD process group, without the need to first create a new group.

Fixes #16922

Before submitting

• Was this discussed/agreed via a GitHub issue? (not for typos and docs)
• Did you read the contributor guideline, Pull Request section?
• Did you make sure 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? (not for typos and docs)
• Did you verify new and existing tests pass locally with your changes?
• Did you list all the breaking changes introduced by this pull request?
• Did you update the CHANGELOG? (not for typos, docs, test updates, or minor internal changes/refactors)

PR review

Anyone in the community is welcome to review the PR.
Before you start reviewing, make sure you have read the review guidelines. In short, see the following bullet-list:

Reviewer checklist

• Is this pull request ready for review? (if not, please submit in draft mode)
• Check that all items from Before submitting are resolved
• Make sure the title is self-explanatory and the description concisely explains the PR
• Add labels and milestones (and optionally projects) to the PR so it can be classified

[codecov]

Codecov Report

Merging #16995 (6a9fa30) into master (7bc2a65) will decrease coverage by 23%.
The diff coverage is 85%.

Additional details and impacted files

@@ Coverage Diff @@ ## master #16995 +/- ## ========================================= - Coverage 82% 59% -23%  ========================================= Files 436 413 -23 Lines 31765 31486 -279 ========================================= - Hits 26009 18687 -7322  - Misses 5756 12799 +7043 

[belerico]

I've seen also that from pytorch 2.0 the torch.distributed.distributed_c10d._pg_map should be called from the _world singleton. I don't know if TorchCollective needs to support something like that, if so one can just add the following

if _TORCH_GREATER_EQUAL_2_0: from torch.distributed.distributed_c10d import _world pg_map = _world.pg_map else: pg_map = dist.distributed_c10d._pg_map

at the beginning and refers to pg_map where needed.
@carmocca @awaelchli, what do you think? I can update the PR in case it's needed