Variance in FuzzBench trials with same random seed

@vanhauser-thc wrote [here](https://github.com/google/fuzzbench/issues/654#issuecomment-672845457):
>the experiment has started => https://www.fuzzbench.com/reports/2020-08-12/index.html
and we can see the aflplusplus_same1 .. aflplusplus_same3 instances to have quite a different coverage.

>for several benchmarks this will even out as their overage saturate after 6 hours, so capturing this now is important.

>most easy to see for curl, re2 and sqlite3:

>|fuzzer|fuzzer1|fuzzer2|fuzzer3|
>|:--:|:--:|:--:|:--:|
>|curl | 17148.0 | 17161.5 | 17155.0|
>|re2| 3461.0 | 3455.0 | 3488.0|
>|sqlite3|26895.0 | 26611.5 | 26420.5|

>as all PRNG values are the same and input selection based on same values this shows that the instances are running with different speeds.