Actively deregister instances from target groups during `eks deploy`

For people using in-cluster mechanisms to manage ELB attachments (e.g., aws-alb-ingress-controller), detaching instances from ASGs do not actually deregister the instances from the ELBs. This can be a source of downtime during the rollout procedure as instances need to fail the health checks before being dropped from the list, which means they can continue to get requests while the servers are being drained and shut down.

Start at the `eks deploy` command and trace its rollout handling for ASGs and ELB target groups. Confirm how instances are detached and determine the target-group deregistration behavior needed before shutdown. Done means instances are actively deregistered during deploy so rollout no longer waits for health checks to remove them.

Written by the indexing model from the issue text.