sql/archiver: keep archive counts using a bucketed list

The buckets are using the last two bytes of the object id, so that we spread the
load across different lines on sequential archivings.