Self-healing stable-egress gateway with automatic failover for the Phoenix server

The Phoenix server’s egress gateway now self-heals when a node is replaced, instead of requiring manual failover. A new leader-elected failover controller (deployed alongside md-egress candidate nodes) keeps the active gateway label and the Hetzner Floating IP on a healthy node, only migrating them when no healthy active node exists, so server egress can no longer black-hole and take the site down on the next node roll. The PR notes that this removes the failure mode behind the 2026-06-14 tuist.dev 503 outage.