Fixing AO tests
Pull Request resolved: https://github.com/pytorch/pytorch/pull/79811
A test was introduced, that was adding a constructor with positional arguments to the instance of the `TestCase`. The intention was to add a reusable "runner" for the relevant testcases. However, this broke the unittest's ability to discover any tests defined.
The failing behavior was not caught because when running the tests using python, the unittest's is able to run the tests correctly, without trying to identify the discovery/listing of all the tests.
The immediate followup is to rewrite the tests in the `data_sparsifier` to implement proper "runner" classes / functions.
Differential Revision: [D37253508](https://our.internmc.facebook.com/intern/diff/D37253508/)
**NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D37253508/)!
Approved by: https://github.com/andrewor14, https://github.com/macandro96
