Making Magic Transit health checks faster and more responsive

Introduced a new distributed health check system for Magic Transit GRE tunnels. This system replaces the previous naive approach where each server independently checked each tunnel every minute. The new system utilizes consistent hashing to assign responsibility for checking specific tunnels to individual servers within a data center. These assigned servers then send health checks more frequently and share aggregated results via multicast heartbeats. This significantly reduced CPU usage by over 70% and memory usage by nearly 85% at the edge.