Rework the scrubber checking data model and cli

Currently, a scrubber storage checking "session" consists in a configuration made of a triplet (datastore "id", object type, partition scheme) -- the partition scheme being simply the number of partitions the hash space is divided into.

This configuration, representing a checking session, is not stored in the DB, but rather depends on the cli user using the same parameters for all the checking workers involved in this checking session.

This scheme does not play very well with elastic deployment (and TBH, neither with standard systemd-based one as well).

A better approach would be to reify this "session configuration" object in the database and make every worker involved in the checking session use this config id instead. To make it easier to use/deploy, the ScrubberDb should also be responsible to "distribute" the partitions to be checked for a given checking session to workers on demand. This would allow to dynamically spawn (and kill) checking workers (thus make it much easier to deploy).

This is needed especially for mirror environments.