RFR: Deflate idle monitors per thread

[Roman Kennke]

Current idle monitor deflation happens when all thread have arrived at
safepoint, but is single-threaded by the VM thread. This patch
implements per-thread idle monitor deflation. This will deflate monitors
when GC calls into Thread::oops_do() or possibly_parallel_oops_do(), and
utilizes the GC worker threads to deflate monitors. This should
parallelize well, and behave better on caches (synchronizer roots are
scanned right after deflation!). All is enabled by
-XX:+DeflateIdleMonitorsPerThread.

http://cr.openjdk.java.net/~rkennke/deflate-per-thread/webrev.00/
<http://cr.openjdk.java.net/%7Erkennke/deflate-per-thread/webrev.00/>


Testing: specjvm, hotspot_gc_shenandoah

Ok?

Roman