Select detection criteria for CI failure of e2e metrics job

[dtfranz]

Look through the data made available from our scraped metrics during e2e, and select criteria that may qualify as CI blocking.

For now, we need only ensure that the metrics collected during the e2e run do not contain any alerts or critical metrics. A critical metric could be, for instance, a count of controller panics greater than zero - in other words, unhandled events that should not happen under any circumstances. These are the kinds of things we want to be looking for.