Let's see if we can't find why these rules aren't generating alerts.

None of these are showing up in the build. It's not yet clear whether
this is a log level / file descriptor issue, or whether our code just
isn't being executed.

I'm not seeing these on CI.

Maybe we can see this?

Understanding what happens during the first rule execution (if there is
one) might help us to understand why we're not generating alerts that
first time.

If the rule is going to eventually succeed, we should see this
eventually resolve if the problem lies in the rule executor's handling
of the failure.

This should give better granularity on the following:

* How long it takes for the ML job to become "started"
* How long it takes for the metrics to become available

See previous commit for context.

There's still a chance that the datafeed/job will _no longer_ be ready
by the time we hit the failing MKI tests (or maybe the timing issue pops
up years from now 😉), but if this makes our tests more consistent
we can start to focus on this: better ML integration.

Let's see how long these are pausing; that might indicate an issue.

Despite our job being started, we're now receiving _no_ alerts when
before we had some. I think this is because the job is starting, but no
anomalies are ready yet. This should validate that hypothesis.

This is less restrictive than the ML helper, which seems to wait for the
job to report as having processed records. Let's see if this
implementation works for us.