Scaling down a MachineDeployment with Machines with the same delete priority may result in too many machines being deleted

[thedadams]

What steps did you take and what happened:

• Deploy a cluster and set the DeletePolicy for all MachineDeployments to Oldest or Newest.
• Scale any MachineDeployment up by >= 2.
• As new machines are provisioning, scale down the MachineDeployment by 1.
  Since all of the currently provisioning machines have the same delete priority as any machine with a non-zero deletion timestamp, this can result in multiple machines (all of the currently provisioning machines) to be deleted.

What did you expect to happen:
Exactly one machine should be deleted when scaling down a MachineDeployment by 1.

Environment:

• Cluster-api version: 1.0.2
• Kubernetes version: (use kubectl version): Tested with 1.21.9, 1.21.10, 1.22.3
• OS (e.g. from /etc/os-release): Any

/kind bug
[One or more /area label. See https://github.com/kubernetes-sigs/cluster-api/labels?q=area for the list of labels]
/area machine

[thedadams]

/assign thedadams

[fabriziopandini]

/milestone v1.2
Nice catch