Weight first request less for dns speed test

The first request should be weighted less. In it no sessions are established yet and it has to be cold-started. This does not reflect how Nebulo normally runs, where it does have sessions and such.

The weighted response time shouldn't be used as min value for the best case measurement though.

Another option would be the ignore the first request entirely for the statistic, always firing n+1 requests. This wouldn't force me to chose some arbitrary weight.