Fixes the issue caused when multile ClusterCIDR objects have the same
nodeSelector values, order of the requirements in the nodeSelector is
not preserved when nodeSelector is marshalled and converted to a string.
Fixes the deletion of ClusterCIDR object, when a Node is associated(has
Pod CIDRs allocated from this ClusterCIDR) with it. Currently the
ClusterCIDR finalizer is never cleaned up as there is no reconciliation
happening after the associated Node has been deleted. This commit fixes
the issue by adding workitems from all events to a worker queue and
reconcile until the delete is successful.