FIX the document of distributed.new_group() (pytorch#122703)

As for now, the document of distributed.new_group() says that it returns `None` when current ranks is not in the new created process group. However, it actually returns `GroupMember.NON_GROUP_MEMBER`. I have check the code and think it is more appropriate that we fix the document.

Pull Request resolved: pytorch#122703
Approved by: https://github.com/wconstab, https://github.com/kwen2501

torch/distributed/distributed_c10d.py

@@ -4100,7 +4100,8 @@ def new_group(ranks=None, timeout=None, backend=None, pg_options=None, use_local
         group_desc (str, optional): a string to describe the process group.
 
     Returns:
-        A handle of distributed group that can be given to collective calls or None if the rank is not part of ``ranks``.
+        A handle of distributed group that can be given to collective calls or
+        GroupMember.NON_GROUP_MEMBER if the rank is not part of ``ranks``.
 
     N.B. use_local_synchronization doesn't work with MPI.