RFR: Use BarrierSetC2::ideal_node() API for CmpPNode::Ideal()

[Roman Kennke]

But the stuff in LoadNode::Identity() is Shenandoah-specific, or is it
not? I don't quickly see how it would fit a generic pre-identity-call
though. So I go ahead with what I just RFRed and you try to upstream the
LoadStore::Identity() stuff? Ok?

Roman

>> Identity() can be treated the same way I think.  Want me to do it?
> 
> We don't need so far so I would say no.
> 
> Roland.
>