RFR (sh/jdk11): Cleanup/revert changes in subnode.cpp

[Roman Kennke]

http://cr.openjdk.java.net/~rkennke/shjdk11-cleanup-subnode/webrev.00/

This part was presumably only needed to get rid of write barriers if
they are no longer use. Problem we had with WB is that the memory
projection would keep them alive even if they had no real use.

Testing: hotspot_gc_shenandoah is fine. I'd like to give CI a chance to
digest it.

Ok?

Roman