Possibly blocked by DM-21877. The issue is being able to test the new runQuantum; for this I need some kind of Gen 3 dataset, but not necessarily the production one (e.g., the Gen 2 PpdbMetricTask uses a fictitious dataset that exists only in the test suite).

Maybe Jim Bosch can comment on exactly how much environment is needed to run runQuantum?

To call runQuantum from test code you'll want to start by making a butler repo in a temporary directory with a SQLite database (possibly an in-memory one), which we do frequently in daf_butler tests (e.g. https://github.com/lsst/daf_butler/blob/master/tests/test_butler.py#L122). You'll generally then need to insert some dimension records to match whatever data IDs you want to use (https://github.com/lsst/daf_butler/blob/master/tests/test_butler.py#L169), and register your dataset types (https://github.com/lsst/daf_butler/blob/master/tests/test_butler.py#L105).  From there you can call put and get.

Hi Ian Sullivan, could you review this? It's 230 lines, mostly test code. (Said test code should be more general than lsst.verify, but generalizing/moving it is for DM-22599.)

Looks good. Just a few minor suggestions on the pull request about formatting and documentation.