Analyze onionperf data for outlier relays during CC deployment

Similar to #12 (closed), we want to see if any particular relays are showing up frequently in bad performance of onionperf data, caused by either the DoS attack, or congestion control.

We may want slightly different kind of analysis as that ticket, because what I want to figure out first is if any particular relays are always showing up in the lower quartile of onionperf performance.

This is lower priority than the Fast and Guard cutoffs in jnewsome/sponsor-61-sims#17, but we may want to use this to filter further relays there before graphing, too.

We will probably want to cross-reference this with sbws vote values too, to see how sbws is voting for these relays.