Extend support of use-cluster-ip to the Ingress resource

[brianehlert]

The option "use-cluster-ip" was released a while back allow compatibility with service meshes. This is because service mesh implementations tend to intercept on behalf of a service cluster ip via modification to iptables or sidecar rules.

The option "use-cluster-ip" is tied to the project CRDs of VirtualServer / TransportServer and is not currently available with Ingress.
There are customers who insist on using the Ingress resource. These customers tend to rely heavily on automation / templating of the Ingress resource as well as the mergeable Ingress model. And because of this express an unwillingness to move to the CRDs because in their situation they cannot granularly set use-cluster-ip only for those Ingress paths where it is necessary.

This is a change to normal NGINX Ingress Controller behavior where the controller process watches the backend service, enumerates the pods, and grants full load balancing control of traffic direct to the pods. This default behavior allows NGINX's intelligent load balancing to dynamically and continually balance the traffic across the pods in an intelligent manner, default to maintaining session congruency, and other built-in NGINX benefits.

[brianehlert]

#4665