Fire stack events as part of `StatusMonitor` state transitions

While working on https://github.com/electric-sql/electric/pull/2781 to fix an issue where we would sometimes omit to fire a `:ready` stack event message when the replication client alone recovered from a disconnection, @alco and I discussed and noticed that the separation between `StatusMonitor`'s state and the stack events can be confusing.

The suggestion here is to essentially move most of the important stack events to be fired from within `StatusMonitor` so that they are consistent with Electric's actual status and behaviour (health check + responses).

This could be in the form of monitoring internal state transitions and firing stack events for particular ones that we already fire events for, but this requires some investigation.