Improper axis limits in histograms

[mkborregaard]

The new Benchmark histograms have the x axis limit set by the edges of the data, but that is not the best for interpretation. Waiting times for a function to run follows a distribution that is bounded at the bottom by 0, and where 0 as a value has a real interpretation. Importantly, the magnitude of variation in the benchmark histogram is best interpreted in terms of the size of the mean value. As such histograms should be bounded on the left by the value 0. On the right it would greatly ease interpretation if they were bounded by a whole power of 10 seconds, e.g. 100 microseconds.