Split the `elasticsearch` sidekiq queue into two dedicated queues
Per https://gitlab.com/gitlab-com/infrastructure/issues/1477#note_27051209 , we're seeing that the elasticsearch queue, which is shared between ElasticIndexerWorker and ElasticCommitIndexerWorker, is a little peculiar.
I think we should split it into two ordinary dedicated queues. This will prevent large git push jobs from backing up issue/MR indexing, and also allow us to throttle the jobs independently. The two use fundamentally different resources, and have very different execution times, so it doesn't make sense to treat them as a unit.