Incorrect reset behavior on some tests

[bvisness]

We have identified two benchmarks that do not reset correctly, leading to benchmark measurements that increase depending on the number of iterations. This is what we are currently using for our benchmark config:

exclude:microBenchmarks.ArrayCopyBenchmark.copyInSameArray
exclude:microBenchmarks.PrimeListBenchmark.calcEratosthenes

There may be others, but these benchmarks have bad behavior due to infinitely-growing arrays.

The problem, it seems, is that the test harness runs the @Setup and @TearDown functions at the beginning and end of the entire test suite, not per iteration. That is, @Setup runs only once before warmups, and @TearDown runs only once after all measurement iterations. This is, frankly, quite counterintuitive and I wonder if that is just a bug in the harness itself.