Reviewed-by: kbarrett, sjohanss
Improve G1 worst-case latencies by making the full GC parallel. Reviewed-by: tschatzl, sangheki, ehelin