Adds more collectives to ProcessGroups (#108)

* initial commit for reduce_scatter

* fixes reduce_scatter function signature, refactors test and adds reduce_scatter test

* fixes test

* adds explicit NotImplementedError to reduce_scatter in gloo, simplify the test suite

* fix tests after merge

* initial commit

* Added all collectives, starting tests now

* fixes allreduce test, add and debug allgather_into_tensor_coalesced

* added all tests, need to hammer out failures

* fixes the tests

* linter

* test simplification

* more cleanup

* refactor + update tests

* add reduce_scatter multi pg tests

* add the full test suite

* final cleanups

* add _assert_list and _maybe_share_tensors