RFR: Simpler obj equals barrier for C2

[Roland Westrelin]

> The next best thing to try is to do resolve_for_obj_equals(a, b)
> approach, right? Or is there a way to get this simpler/better(IMO) API
> to work without too much messing about?

One case where using the simpler API is complicated is with some
intrinsics that have predicates. The code that emits the predicates only
return some control but it would now need to return control + new memory
state and merge the new memory state. So several methods in shared code
would have to be modified.

Roland.