Spencer Nelson mentioned that files can be marked as generated to keep them from appearing in GitHub diffs. This would be very useful for the ap_verify datasets, whose updates include some human-unreadable text files.

In ap_verify_testdata, ap_verify_ci_cosmos_pdr2, ap_verify_ci_hits2015, and ap_verify_hits2015, mark at least the export.yaml file as generated (since this is always a dump of the "preloaded" repository). Consider also marking gen3.sqlite3 and other Gen 3 repository files; however, at least the last would need to be undone once we moved to pure Gen 3 datasets.