Figure out the methodology for native benchmarks about rsc

[xeno-by]

From the point of view of run-to-run variance, I think that existing methodology has proven to be quite useful. Here's a summary of 10 cold and 10 hot native typecheck runs (showing that variance seems to be on par with JMH: #33):

ColdRscNativeTypecheck.run   cli  100  299.179  ms/op
ColdRscNativeTypecheck.run   cli  100  297.612  ms/op
ColdRscNativeTypecheck.run   cli  100  298.294  ms/op
ColdRscNativeTypecheck.run   cli  100  297.597  ms/op
ColdRscNativeTypecheck.run   cli  100  297.936  ms/op
ColdRscNativeTypecheck.run   cli  100  297.955  ms/op
ColdRscNativeTypecheck.run   cli  100  298.402  ms/op
ColdRscNativeTypecheck.run   cli  100  298.633  ms/op
ColdRscNativeTypecheck.run   cli  100  300.401  ms/op
ColdRscNativeTypecheck.run   cli  100  300.134  ms/op
HotRscNativeTypecheck.run   cli    1  245.171  ms/op
HotRscNativeTypecheck.run   cli    1  241.071  ms/op
HotRscNativeTypecheck.run   cli    1  240.934  ms/op
HotRscNativeTypecheck.run   cli    1  242.211  ms/op
HotRscNativeTypecheck.run   cli    1  243.172  ms/op
HotRscNativeTypecheck.run   cli    1  241.510  ms/op
HotRscNativeTypecheck.run   cli    1  241.346  ms/op
HotRscNativeTypecheck.run   cli    1  246.583  ms/op
HotRscNativeTypecheck.run   cli    1  241.673  ms/op
HotRscNativeTypecheck.run   cli    1  241.312  ms/op

In order to not spend time on splitting hairs, I'll be considering this issue completed, unless there are better suggestions that are similarly fast and easy to implement. Everyone, please feel free to chime in.