I've got a few weeks worth of data for runs=5 from which
I can check how often time-based values had false negatives.
Let's see if runs=7 performs better (more evenly distributed
between faster and slower), or worse (more likely to get a
super slow result that is otherwise absent).
Change-Id: I38e9b4c2c3c2f4832fc0e9244a00a723510d295f
warmup: true
-runs: 5
+runs: 7
scenarios:
Read a page:
# The only page that exists by default is the main page.