fix: cdc tenant cache usage

[edgurgel]

What kind of change does this PR introduce?

• Changes so that the Postgres Changes pipeline for a tenant restarts on a database change
• Changes to that the Postgres Changes pipeline uses the cache instead of holding the tenant information in memory. This ensures that

What is the current behavior?

Postgres Changes pipeline might have the old tenant definition for a while until the pipeline stops or crashes

What is the new behavior?

Once the database address is updated we tell the pipeline to stop and once resubscribed (this is automatic from the RealtimeChannel system) the new tenant information is used to reconnect.

Even if the pipeline failed to restart the pipeline will eventually have connectivity issues and naturally restart due to the supervisor behaviour. Once it restarts it will fetch from the cache again which should have the new information (either from the invalidation or from refetching after expiring)

Additional context

Add any other context or screenshots.

[coveralls]

coverage: 88.531% (+0.2%) from 88.352%
when pulling 1865d1e on fix/cdc-tenant-cache
into f13d79d on main.