RFR: (sh/jdk11): Move has_object_fields() from shared to shenandoah

[Roman Kennke]

In order to clean up Shenandoah vs upstream diff a little more, I 
propose to move the has_object_fields() stuff from ciInstanceKlass to 
under Shenandoah. It has the downside that it cannot be 
recomputed+cached, but I believe this is a minor issue. We should only 
get there when considering (non-expanded) clones, and only for non-array 
types, it should be rare enough and performant enough to not cause troubles.


http://cr.openjdk.java.net/~rkennke/shjdk11-hasobjfields/webrev.01/

Testing: hotspot_gc_shenandoah

Can I please get a review?

Roman