- currently, if we don't recognize the same diff several times (e.g.
if you restart the storage diff watcher pointed at the same file),
we'll add the same row to the queue on each run.
- these changes assure we only queue an unrecognized diff once.
libraries/shared; adjust omni test_helpers to drop and recreate
checked_headers table to avoid reaching 1600 column limit after repeated
tests (dropping columns doesn't actually remove them from contributing
to that limit)
