Use Median instead of Mean whe deducing Overhead

[adamsitnik]

I just had a conversation with @AndyAyersMS and Andy asked if we remove the outliers from the overhead.

I did check our code and it looks like we don't:

BenchmarkDotNet/src/BenchmarkDotNet/Engines/RunResults.cs

Line 40 in 52eca7a

double overhead = Overhead == null ? 0.0 : new Statistics(Overhead.Select(m => m.Nanoseconds)).Mean;

Andy has run into a case where it would make a lot of sense:

1.69
1.65
1.70
1.83
2.60 <-- an outlier -->
1.76
1.69
1.69
1.73
1.64
1.66
1.67
1.71
1.68
1.64

@AndreyAkinshin I think that we should remove the outliers from the Overhead. Do you agree?

[AndreyAkinshin]

@adamsitnik thanks for report, it's a good point. However, I decided that it will be better to just use Median instead of Mean.

[adamsitnik]

Median instead of Mean

the simplest solutions are always the best! thanks!