Implement delayed termination to achieve zero downtime upgrades

[pleshakov]

To address this, a common approach is to introduce delay in pod termination to allow enough time for an external load balancer to drain its upstream servers.

Acceptance criteria:

• Introduce configurable delayed termination.
• Rerun zero downtime non-functional tests and update results for 1.0

More info:

• https://kubernetes.io/docs/reference/networking/virtual-ips/#traffic-to-terminating-endpoints
• https://github.com/kubernetes/enhancements/tree/master/keps/sig-network/1669-proxy-terminating-endpoints
• https://learnk8s.io/graceful-shutdown