Please track durations for key moments of our jobs.  Some initial ideas

1. overall install duration
2. time required to provision masters during install
3. time required to provision workers during install
4. install cluster bootstrap duration
5. overall upgrade duration
6. upgrade duration prior to node updates
7. overall job duration
8. duration of openshift-tests

Getting data for some within the next couple months will help us have a historical basis when we push to 4.17.

Complication here is that origin is not running during install. The resource watch observer however is, and is generating intervals and other artifacts, just not the ones we typically display or work with. Could the intervals be generated from artifacts in the cluster itself after tests run? (seems unlikely we could find everything we need in events/pod logs)

Intervals would be nice to see displayed for this, but autodl json files might be better for graphing options and reducing the amount of data we'd have query to do useful things with it. Both might be optimal.