Using the tools created for WT-6662, et al., run a test to measure the performance impact of history store growth in a degraded PSA replica set and present the results here.

The goal is to describe the basic test and the numbers I get when running a modest size instance of the test.  Based on feedback we can adjust what the test does and expend the data set and runtime to potentially stress the system more.

Because this ticket is intended as an exposition of results, I am revising my comments as I figure things out, rather than having a string of comments that correct and clarify previous ones. 

[Update: I was misinterpreting results from Genny—measurements I thought were in microseconds were really in nanoseconds.  Since the comments here are more of a report than a conversation, I'm going to go back and edit them to present the correct values.]