Reviewed-by: ehelin, coleenp, kvn, ihse
Improve G1 worst-case latencies by making the full GC parallel. Reviewed-by: tschatzl, sangheki, ehelin