Request for Review (xs) - 8156867: Simplify/reduce testing in ParallelCompact_test

[Jon Masamitsu]

This is a bug fix for after the FC snapshot.

https://bugs.openjdk.java.net/browse/JDK-8156867

The behavior for which the ParallelCompact_test was written
was that the logging accessed the end ParallelGC region (which
should not be used), found junk there, and failed an assertion.
The fix was to prevent the logging from accessing the end ParallelGC
region.  The test put junk in the end ParallelGC region and then
called the logging (which should not have accessed the end ParallelGC
region but, if it did, it would find junk there and fail).  The end 
ParallelGC
region again should not be used.  On some platforms it seems to be allocated
but not on others.  When the latter was the case, the segv occurred.
I deleted the setting of junk in the end ParallelGC region.  Now if the
test does access the end ParallelGC region it will fail if there is the
right junk in it.

Change deletes 7 lines.

http://cr.openjdk.java.net/~jmasa/8156867/webrev.00/

Tested locally on a machine where the segv was happening
and on one where it was not.

Thanks.

Jon