Request for Review (xs): 8151101 - Improve UseParallelGC parallelization of object array processing

[Jon Masamitsu]

This is a contribution from richard.reingruber at sap.com

When long object arrays are processed, a GC worker will take
the front of the array to process and push the remained of
the array onto a stack to make it available to other GC workers.
The current code does "process my part of array" and then
"push the remainder".  This changes that to "push remainder" and
then "process my part of array".

See CR for performance comparison on micro benchmark.

https://bugs.openjdk.java.net/browse/JDK-8151101

http://cr.openjdk.java.net/~goetz/wr16/8151101_paroldgc_parallelization_of_large_obj_arrays/webrev/

Thanks.

Jon