Somehow include failed builds in the `cloud_summary()` and `cloud_broken()`

[jimhester]

Unfortunately sometimes AWS preempts jobs and kills them, which results in uninformative failures.

Compounding this problem is currently we don't include these failures in the cloud_summary() or cloud_report(), so depending on which jobs failed you might not know a regression occurred.

As the only thing we can really do is re-run these jobs we need to have a function to query which ones failed.

[jimhester]

I will work on this in the next few days, but wanted to have an issue to track it.