Update import job to send unique timestamp and import_entity_table_check event

[ericaporter]

The current DfE Analytics import job (aka Backfill) in ruby imports all BQ entities to BQ. The job needs further enhancements to allow DfE Analytics to do a reliable sync of a complete table ie update to the latest data for each row and delete any extraneous rows in BQ.

This PR includes the following changes:

1. Adds a single unique import entity id to ALL events for a given import as a numerical timestamp in the format: YYYYMMDDHHMMSS. The field for this in the event could be event_tag

2. For each entity, once the import has been completed, a new checksum event is sent (import_entity_table_check)
   which contains the unique entity import id to the event_tag

3. A checksum event can now be ordered by ID if created_at / updated_at are missing. Note that this changed will also affect the nightly checksum job

4. The entity table check and checksum calculation logic has been refactored into service objects

[asatwal]

This is looking really good @ericaporter . Don't think you need to turn any other of the other smaller modules into services. Just a few more simple collaboration specs I think ?

[asatwal]

I thought maybe ID should be in ALLOWED_ORDER_COLUMNS, but then the where clause would be wrong if that was done. Maybe ALLOWED_ORDER_COLUMNS should be renamed to something like BUILD_WHERE_CLAUSE_FOR_ORDER_COLUMNS.

[asatwal]

The collaboration specs for the services are a nice to have.

[asatwal]

This is looking good @ericaporter - With hindsight I can see why PG gets confused by the ID column ordering in the Checksum subquery. Well done for finding the problem and persisting with this.