Frequent health checks generate constant background CPU usage

Istio runs active health checks against upstream endpoints. Default interval is 10s. With many services and endpoints in the mesh, continuous health checking generates constant background CPU usage that scales with endpoint count. This is independent of actual traffic volume.

Increase outlierDetection interval in DestinationRule from default 10s to 30s. This reduces CPU usage from health checking by roughly 3x. Also set consecutive5xxErrors: 5 and baseEjectionTime: 60s to maintain reasonable failure detection. The tradeoff is slightly slower detection of unhealthy endpoints (30s vs 10s).