[11u] RFR 8249192: MonitorInfo stores raw oops across safepoints

[Aleksey Shipilev]

Original bug:
   https://bugs.openjdk.java.net/browse/JDK-8249192
   https://hg.openjdk.java.net/jdk/jdk/rev/61f9028f360d

The bulk of the patch applies exactly, but some hunks need adjustments.

  *) In jvmtiEnvBase.cpp, the hunk removing "ResourceMark rm(current_thread)" cannot be applied, 
because "current_thread" is added with JDK-8247729:
   https://hg.openjdk.java.net/jdk/jdk/rev/f8a9be0f9e1a#l2.82

  ...I just removed the ResourceMark by hand.

  *) In deoptimization.cpp, there is no eliminate_locks yet -- it is refactored out of "if 
(EliminateLocks)" branch later with JDK-8226705:
   https://hg.openjdk.java.net/jdk/jdk/rev/408c445d04e8#l21.162
   https://hg.openjdk.java.net/jdk/jdk/rev/408c445d04e8#l21.59

  ...so I put the HandleMark in the old place.

Note it has a follow-up JDK-8251118, which needs to be backported too. Since the whole thing has a 
non-zero risk, it targets 11.0.10.

11u webrev:
   https://cr.openjdk.java.net/~shade/8249192/webrev.11u.01/

Testing: tier{1,2,3} with and without Shenandoah [which seems to find some of these bugs]

-- 
Thanks,
-Aleksey