Make the hashed storage migration rake tasks not depend on sidekiq-queue size

Follow up from: #995 (closed)

The hashed storage migration can be triggered from rake tasks. Those migrations guard against starting rollbacks/migrations simultaneously by checking the queue size of the respective workers.

These runners aren't used on GitLab.com anymore, so these don't block the migration of catchall to a limited set of queues in &447.

But for self-managed installations we need to keep these workers on their own queue to keep this working the same way. Should we figure out a way to make these rake tasks not depend on the queue size?