gs_offloader has been (plaintively) complaining about offload
failures since (roughly) mid-December.  The noise has been getting
progressively louder; it's now the single biggest source of complaint
noise on the chromeos-infra-alerts alias.

A spot-check says that the offload complaints are about specific
directories that never offload.  We should go figure out why they
fail and how to stop such failures in future.  We should also then
either delete the problem directories, or get them safely offloaded.

Here the most recent sample complaints from
chromeos-server11.hot.corp.google.com:

First failure       Count   Directory name
=================== ======  ==============================
2017-01-13 15:21:26    117  92202181-chromeos-test
2017-01-13 15:21:28    116  hosts/chromeos4-row9-jetstream-host5/59253162-repair
2017-01-13 15:21:30    116
hosts/chromeos4-row10-jetstream-host7/59385679-repair
2017-01-13 15:21:32    116
hosts/chromeos4-row9-jetstream-host5/59252950-provision
2017-01-13 15:21:33    116
hosts/chromeos4-row9-jetstream-host3/59480989-provision

For comparison, complaints from cros-autotest-shard1.cbf.corp.google.com:
First failure       Count   Directory name
=================== ======  ==============================
2017-01-13 15:21:53    117
hosts/chromeos4-row10-jetstream-host5/59385677-repair
2017-01-13 15:22:00    117
hosts/chromeos4-row9-jetstream-host4/59308759-provision

The times I've spot-checked failures, complaints from
jetstream hosts and tests have been common.