Sylvain Henry authored Jan 22, 2021 and Marge Bot committed Jan 27, 2021

* disable idle GC which has a big impact on time measures

* use average measures (before and after event registration)

* use warmup measures (for some reason the first measure of a batch
  seems to be often quite different from the others)

* drop the division by monotonic clock time: this clock is impacted by
  the load of the runner.  We only want to measure the time spent in the
  RTS while the mutator is idle so I don't understand why it was used.