kube-controller-manager doesn't check for resource presence when deleting legacy providers

    kube-controller-manager operator is cleaning up resources for legacy in-tree controllers, running multiple delete requests to apiserver. 
This has several drawbacks:
* API Server can't cache DELETE requests, so it has to request etcd every time
* Audit log is flooded with "delete failed: object 'foo' not found" record
The function should first check that the object exists (GET requests are cachable) before issuing a DELETE request

    1.
    2.
    3.