Additional logging to support failure metrics

As a product manager/tech lead I want to know what percent of total Travis PR build failures are caused by flaky tests over a specific time period so I can determine if flaky tests are having a significant impact or not. 

Count of PRs with flaky failures over X period / Count of total PRs over X period