RFR: 8221207: Redo JDK-8218446 - SuspendAtExit hangs

[David Holmes]

Bug: https://bugs.openjdk.java.net/browse/JDK-8221207

There was one small oversight in the original fix that led to crashes, 
seen (randomly) in JDI tests. The safepoint check must not happen if the 
thread-state is already _thread_in_native. I've checked the thread-state 
on all call paths to confirm that.

Incremental webrev from original fix: 
http://cr.openjdk.java.net/~dholmes/8221207/webrev.inc/

Full webrev: http://cr.openjdk.java.net/~dholmes/8221207/webrev/

Re-tested in mach5 tiers 1-3 and com/sun/jdi tests (but they passed last 
time too.).

Thanks,
David