Ingress maintenance

Use routerctl ingress drain when an IngressService backend needs temporary
maintenance without editing the router YAML:
routerctl ingress drain ingress/kubernetes-api backend=cp-01 --duration 10m
routerctl get IngressService
The drain state is stored in the routerd state database. During the drain
window, the ingress controller marks that backend as drained: true,
healthy: false, and reason: Drained; new flows are sent to the remaining
healthy backends on the next reconcile. Existing conntrack entries are not
flushed.
The backend is restored automatically when --duration expires. To restore it
immediately:
routerctl ingress undrain ingress/kubernetes-api backend=cp-01