Optimize indices_to_evict_check with a single batch

Background

indices_to_evict_check runs on full records of zoekt_indices. This can put too much stress on the database. The second issue could be that in a single run, all indices might not get processed and they will get re-enqueue after 10 minutes which will increase redundancy.

Proposal

Publish the event in the scheduling_service without any indices_id. In the worker IndexToEvictEventWorker add order(:id) and limit(1000) to process just one batch.