Related T494

Related T494

This will allow us to handle another storage backend for the storage of the
archiver data.

Permits to batch update content rows (with or without optional new
columns).  Limited to contents (table content only, table
skipped_content is not dealt with).

Related T692
Closes D185

By default we would try to copy objects from all the archives, even those for
which we didn't have a configuration.

The default value for content copies is "missing", so we don't need to make it
explicit.

Implementation wise, this uses COPY statement and drop duplicates if
encountered for content_archive insertion.

Use the same insertion pattern as swh.storage.content_add.

This also has the benefit to hide some celery name (which is an
implementation detail from swh.scheduler).

Related: T494

This use case cannot happen with ArchiverWithRetentionPolicyDirector:
- If a row entry is referenced in the archiver db, it's present in
  the objstorage

- And if a row entry is not referenced in the archiver db, it
  won't be listed as missing since it's the archiver db which is
  read for listing the contents we want to archive.

Related T494
Related T569

Related T494

To make sure corruptions such as T680 don't happen again, use the same
normalization function as swh.model before inserting timestamps into our
database.

This makes swh.storage reject non-integer timestamp values as well.

Update tests to reflect this change.

Close T672

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

By default, return data from the latest visit instead of returning data from all
visits, which doesn't make much sense.