Optimize create_collections and get_or_create_collections

[data-sync-user]

We have a shared constant FIRST_CUSTOM_COLLECTION_ID that is used to define collections >= 101 as being custom. We have defined the standard static collections but need a mechanism to ensure added collections insert past the custom value. Spanner utilized this in this create_collection calls. We addressed this in our database migrations explicitly, but this pointed out that we should refine the way these methods work so we minimize needless calls.

MySQL does not do this and simply used a database migration entry to define the next value. We want to make sure we address this in Postgres and make it closer to the Spanner implementation.

┆Issue is synchronized with this Jira Task