Results of individual benchmarks are affected by the amount of benchmarks in the suite

I reproducibly get different results from the same benchmarks depending on whether I include other benchmarks in the suite.

These are the results I get, [when the suite contains other benchmarks](https://github.com/nikita-volkov/folding.elm/blob/af70d0f0950e5e6d81c67b7086a4683017e4407f/benchmarks/Unfoldl.elm):

And these are [when the suite contains these benchmarks only](https://github.com/nikita-volkov/folding.elm/blob/05a2a7f4fb35075c64639e7b415a58fd0064f8f5/benchmarks/Unfoldl.elm):