RFR: Cherry-pick/synchronize JDK-8180175 and JDK-8180599"

[Roman Kennke]

I've upstreamed the fast synchronizer root scanning code (i.e. scan only
thread-local monitor lists). It turned out to introduce a bug where we
could miss monitors on thread termination, and also moved code in
different places. This change synchronizes the corresponding code with
what I upstreamed.

http://cr.openjdk.java.net/~rkennke/fastsync/webrev.00/
<http://cr.openjdk.java.net/%7Erkennke/fastsync/webrev.00/>

Testing: hotspot_gc_shenandoah

Ok?