Description of problem:
Running hack/local-up-cluster.sh with `KEEP_TERMINATED_POD_VOLUMES=false`, start a busybox pod that terminates after 10s and does not restart:

```
apiVersion: v1
kind: Pod
metadata:
  name: busybox
spec:
  containers:
  - name: busybox
    image: busybox
    command:
    - sleep
    - "10"
  terminationGracePeriodSeconds: 0
  restartPolicy: Never
```

In ~10s, the pod goes to Complete.  2 minutes later two events occur:

```
2017-07-26 16:32:54 -0500 CDT   2017-07-26 16:32:54 -0500 CDT   1         busybox   Pod                 Warning   FailedMount   kubelet, 127.0.0.1   Unable to mount volumes for pod "busybox_default(aaa4c2ce-7249-11e7-8f15-7085c20cf2ab)": timeout expired waiting for volumes to attach/mount for pod "default"/"busybox". list of unattached/unmounted volumes=[default-token-3dgs5]
2017-07-26 16:32:54 -0500 CDT   2017-07-26 16:32:54 -0500 CDT   1         busybox   Pod                 Warning   FailedSync   kubelet, 127.0.0.1   Error syncing pod
```

The occurs because a `syncPod()` is called from a pod worker after the volumes have already been unmounted, because the pod is terminated, and `kl.volumeManager.WaitForAttachAndMount(pod)` hangs for 2 minutes then times out as the volume reconciler will not remount these volumes.


Version-Release number of selected component (if applicable):
3.6.0

How reproducible:
Always

Steps to Reproduce:
1. see above
2.
3.

Actual results:
2 spurious events 2 minutes after pod termination

Expected results:
No spurious events from terminated pods

Additional info:

This is already fixed in kube upstream:
https://github.com/kubernetes/kubernetes/pull/49724

And origin master:
https://github.com/openshift/origin/pull/15533

This bug is track a backport to OCP 3.6.z