taint "node.cloudprovider.kubernetes.io/uninitialized=true:NoSchedule" is applied to new infra node because azure-cloud-node-manager pods doesn't have right toleration. Fix of new version is able to match key: node-role.kubernetes.io/infra but not key: infra.

    4.14.z

    1.  I modified taints in my test infra node machine set from  

- effect: NoExecute
          key: infra
          value: reserved

to 

- effect: NoExecute
          key: node-role.kubernetes.io/infra
          value: reserved

2. Then I created a new infra node using the previous machine set.
3. In this case new infra node has been created, taint "node.cloudprovider.kubernetes.io/uninitialized=true:NoSchedule" has been removed, new infra node has been added in backend pool of the internal load balancer, and azure-cloud-node-manager pod is running on new infra node.     
    

    https://issues.redhat.com/browse/OCPBUGS-34556 Fix of new version is able to match key: node-role.kubernetes.io/infra but not key: infra.

     https://issues.redhat.com/browse/OCPBUGS-34556 fix of this should be able to match key: infra