This waits for DiskPressure to no longer exist before pre pulling images
after a DiskPressure test. Otherwise we risk to pull images, while
kubelet evicts/removes them.
The "system_node_critical_test" causes DiskPressure on the node,
resulting in eviction of some of the pre pulled images. This makes all
the resulting tests to fail, since their pod spec use PullPolicy: Never
The PrePullImages is now inside a defer, so it will be executed even tho
the assertions inside the "AfterEach" fail.