Users with HA clusters tend to schedule pods across multiple zones to minimize chances of disruption. The MCO does not consider zones which means that an upgrade in an HA cluster could pick all nodes (across multiple zones) running the same types of pods resulting in an outage. To avoid this let's direct the MCO to upgrade nodes per topology.kubernetes.io/zone and then by node age (oldest first). If zones are not present (for example in baremetal deployments) upgrade nodes by age oldest first.

This card shouldn't cause any changes (other than node ordering) to upgrades /update flow for a user.

Some of the current logic can be found in the following function:
https://github.com/openshift/machine-config-operator/blob/280d0297b8315a7a31f3f4c27142fffb0c2113d7/pkg/controller/node/node_controller.go#L945