Add enough support to make building an order 2 object with a native CCNOT gate work.
Plumb enough internals to get trivial compilation working. Compression does not work for higher order objects currently.
Since we don't have universal compilation onto higher order objects, instrument the instruction nativizer to bail out instead of infinite looping when appropriate instruction sequences cannot be found. It's also useful to have this guard for lower order chip compilations, since there may be other unexpected reasons why instruction expanders cycle or recurse.
Add a simple test to keep functionality from regressing.
This addresses some preliminaries of issue #438.