* use 2x chrome total time as benchmark
* set TLS minVersion to v1 only if connection already failed with TLS version mismatch
* correct percentile function
* assert at least 1000 requests were made
* setNoDelay on HTTPS-over-HTTPS requests
* allow for tests with HTTPS upstreams to be slightly slower
* try 3x
* add note for add'l multiplier on httpsUpstreamProxy