add index by user_id on classifications

[yuenmichelle1]

Add index by user_id on classification_events. Currently, we are indexing by event_time as well as our composite primary key which automatically creates index by pkey (a combo of event_time and classification_id).

Originally, classification_events was designed without the index by user_id mainly because

• ERAS' responsibility is to be a classification/comment counter (of when and how many)
• ERAS leaves the responsibility of counting by user_id to the materialized view/continuous aggregate. (DailyClassificationCountsPerUser____)

As we look into sync script (syncing newly added group members and their classifications), we now care about the case where users who are supposed to belong to a user_group but did not join the user_group (and have classifications that should be counted to the user_group). And therefore we need to search classification_events by user_id. This search is currently inefficient since it has to go through sequential scans of the whole hypertable in order to receive results.

NOTE
-Timescale currently does not support creating indexes concurrently. This means that we cannot avoid write locks.

• HOWEVER, Timescale DOES offer adding indexes on a transaction per chunk basis which extends postgres' CREATE INDEX
  See: https://docs.timescale.com/api/latest/hypertable/create_index/ for more details.

[yuenmichelle1]

@zwolf @lcjohnso as long as we are encountering this, is there a need to add indexes for any other tables of ERAS?

(My gut says no, but I also am being conservative since indexes take up storage space)

@lcjohnso As I keep perspective, this issue of seq scans when searching by user_id will be an issue for the sync job not just for the first time run, but also for subsequent runs until we optimize the query of searching classification events by specific user_ids (Hopefully this change should optimize the query).