use performMajorGC

In 1.4.0.0 the `performGC` was changed to `performMinorGC`

> * Use `performMinorGC` rather than `performGC` to update garbage collection
statistics. This improves the benchmark performance of fast functions on large
objects.

While that is true, it also "destroys" accuracy of a bit slower functions. In case when `measure` (which allocates relatively big `Measure` object on each iteration, and there are enough iterations, triggers a GC which must be major: that run is a massive outlier.

IMHO, it's simpler just to accept that GC have to be run, this is not right place to optimize for performance.

Start by locating measure and its current performMinorGC call, then inspect how garbage-collection statistics are gathered for each iteration. Done means the benchmark accounts for major collections again and no longer hides the large outlier described in the issue; use the existing benchmark or measurement checks to verify the behavior.

Written by the indexing model from the issue text.