[cassandra] POC the cross replication pg<->cassandra infrastructure in staging

In order to test the feasibility of the cross replication pg<->cassandra in a pseudo real environment, we should deploy the poc in staging.

Target:

• Kafka -> cassandra replayers (already done to populate the cassandra backend)
• Kafka -> postgresql replayers (to insert in postgresql the objects loaded by the workers using cassandra as a backend)

TODO:

• Extend the replayers helm configuration to support postgresql configuration and a filtered configuration with a retry and a filter step
• Create a new privileged kafka user (https://docs.softwareheritage.org/sysadm/deployment/howto-add-journal-user-credential.html#deployment-howto-add-journal-user-credential)
• Bootstrap the consumer groups to start at the end of the topics as postgresql is already up to date
• Deploy postgresql replayers
• Possibly deploy new nodes or resize the current ones to handle the new load (should not be necessary at the beginning)
• Monitor the behavior of the workers and replayers (not too much errors in sentry regarding the HashCollisions / Exception / slowness)
• For staging, test with more workers to check on load