Implement detection of broken Beaker hosts

[spbnick]

Sometimes test runs fail because there is an issue with a host in Beaker. We need to detect such conditions to avoid sending such failures to developers.

One way to do that would be running tests with a known-good kernel on the same host after a failure is detected. If that fails as well, we can assume that the host is at fault. Otherwise, we can assume that it's a genuine test failure.

We can implement the above by passing a URL of a known-good kernel to skt's Beaker runner, and the known-good kernel can be the baseline the patches are applied to. Sktm would need to be in control of, or at least know, the kernel locations before it can do that.

Also check the (now non-)existing logic in case there are some extra smarts we can use.