Make the archiver use content_archive.num_present

content_archive.num_present act like a cache for the number of copies
where the content is present. This allow to make a much faster request
when retrieving the contents that have fewer copies than expected.