Update flush LSN based on keep-alive message

As per a comment on Twitter, Postgres keep-alive messages contain an LSN which we may use to update the flush LSN and make progress with the replication slot, also if there's no traffic, and without the need to induce artificial traffic via heartbeat action queries. This would help to mitigate the low-traffic/high-traffic scenario potentially causing WAL pile-up when connecting the replication slot to the low-traffic DB. Also see this related change in psycopg2 as well as this bugfix.