Fix database compaction crash on retry If the compaction is retried 2 or more times, there's a badmatch when asserting that the number of changes processed equals the number of total changes. This is due to the fact that the task properties were not updated properly once a retry happens.