RFR: Refactor alive-closures to deal better with new marking contexts

[Roman Kennke]

Turns out this was easier to do than I suspected. RefProc does already
take NULL alive-closure and resets it when used. This means that we
already carried unnecessary cruft in the form of our static
alive-closure and forwarded-alive-closure. The tricky part was to
abstract out the selection logic, this is now in a little helper class
called ShenandoahIsAlive.

http://cr.openjdk.java.net/~rkennke/aliveclosures/webrev.00/

Testing: tier1_gc_shenandoah passed, pending tier3_gc_shenandoah before push

WDYT?

Roman