Logic for priority reindexing

As described in the parent epic there are certain scenarios where we want to be able to rebuild an indexes in a more targeted fashion on GitLab.com: For example, we might have identified a pathology in #218410 (comment 571485758) and want to re-create the index. Another reason to do a targeted rebuild is an INVALID index left behind and similar situations.

This issue focuses on creating priority reindexing heuristic. If there is an entry in the queue, then it takes priority. If the queue is empty then default to using the bloat heuristic.