Repeated test mode

[tollmanz]

I ran into an issue today where I was reviewing timelines and found that the evaluation time for a piece of JS varied rather dramatically. I had to repeatedly generate timelines until I could find what seemed to be a representative result. This work can and should be automated.

I'd like to introduce a "repeated test mode" where you instruct the tool to generate x number of timelines and then instruct it to return the one that satisfies some criteria (e.g., median first-meaningful paint). A first pass at this would keep the criteria limited to relatively simple metrics. A further step could be to pass a script that does more complicated evaluation of the timeline to choose.