Configure a separate queue for batch reindexing jobs

Executing a full reindexing sends the Sphinx indexation jobs to the "default" sidekiq queue. This may delay the processing of some of the other jobs that are also using the "default" queue, and also can almost block the lower-priority queues.

The task is to split these reindexing jobs into a separate queue, with lower priority. This would give us two main benefits:

• the reindexing jobs will be executed with lower priority so will not affect other more time-critical jobs
• it will be possible to delete all reindexing jobs in case something goes wrong

NOTE: this should not affect the indexing jobs happening in real-time on indexed objects updates.