Pod culler needs to be more robust

[betatim]

pod-culler crashes if a pod has no start time

$ kubectl logs pod-culler-96bb965fd-sgwf8
[I 2018-04-03 15:54:32 culler:88] watching for pods with label selector component=singleuser-server in namespace prod
[I 2018-04-03 15:54:32 culler:143] Culling pods older than 06:00:00
Traceback (most recent call last):
  File "/usr/local/bin/culler", line 233, in <module>
    asyncio.get_event_loop().run_until_complete(main())
  File "/usr/local/lib/python3.6/asyncio/base_events.py", line 467, in run_until_complete
    return future.result()
  File "/usr/local/bin/culler", line 228, in main
    culler.cull()
  File "/usr/local/bin/culler", line 151, in cull
    age = now - pod.status.start_time.replace(tzinfo=timezone.utc)
AttributeError: 'NoneType' object has no attribute 'replace'