RFR: Assert Shenandoah-specific safepoints instead of generic ones

[Roman Kennke]

Am 27.11.2017 um 15:43 schrieb Aleksey Shipilev:
> http://cr.openjdk.java.net/~shade/shenandoah/assert-shenandoah-sps/webrev.01/
> 
> There are intermittent assertion failures in ShenandoahHeap::assert_gc_workers. It happens almost
> always when "Concurrent cleanup" is running in the middle of biased locking revocation safepoint.
> Then, assertion code thinks it should run with UseParallelGC threads.
> 
> Fixed that by checking for Shenandoah-specific safepoints only. Other asserts are changed to verify
> all Shenandoah-specific safepoints are covered.
> 
> Testing: hotspot_gc_shenandoah (fastdebug)
> 
> Thanks,
> -Aleksey
> 

Looks reasonable.

Roman