Pod hubs wait for the cluster before announcing themselves

Some of the portal's central message hubs are created very early during startup. They used to announce themselves to the cluster immediately, even though the starting portal had not yet become an active cluster member. During a rolling update that could leave no active portal able to accept the announcement, producing a burst of routing errors and a short interval in which those hubs could not be reached from another portal.

Those announcements now wait for the same lifecycle milestone as the portal's other cluster routing setup. Once the portal is active, both routing paths start normally. A hub removed before that point no longer sends a redundant withdrawal for an announcement it never made.