move health checks back into job definition

[tgross]

This PR eliminates the separate checks package and moves health checks back under the job definition.

Included:

• changing of terminology poll and frequency to interval in all user-facing configs (I've left these separate internally for clarity when they're side-by-side in the code
• made log entries more consistent between watches, jobs, and sensors.
• test_tasks in the integration tests was flaky because the timeout of one of tests was very close to its frequency (now "interval"), and the test has been constrained to eliminate this flakiness. An end-user that makes the timeout and interval very close together will not be able to guarantee a specific number of restarts if that field is set. This is an inherent limitation in handling the timeouts and restarts asynchronously, but it's also not really a problem under realistic use. We'll document it though.
• a bunch of test and test fixture cleanup

cc @misterbisson @jasonpincin @geek @cheapRoc

[tgross]

I'll push up a PR matching these requirements to RFD86 later today or tomorrow morning.

[tgross]

Pushed up TritonDataCenter/rfd#32