API does two DB accesses when getting node, port

The API sometimes does two DB accesses when getting node and port (and portgroup) objects. This is a performance issue when getting thousands of nodes. (Noticed by Yahoo.)

The problem is that our nodes have a chassis ID. The API though, only exposes the chassis UUID so it will get the chassis object in order to get the UUID:

https://github.com/openstack/ironic/blob/edc37cbe1db539372a5e96296d7fbbebe9c50b2f/ironic/api/controllers/v1/node.py#L596

Jim Rollenhagen said that he has addressed this with his port changes:
https://review.openstack.org/#/q/status:open+project:openstack/ironic+branch:master+topic:refactor-object-registry

(I'm guessing https://review.openstack.org/#/c/262098/, but I haven't looked at them yet.)

Similarly, our ports and portgroups have node IDs.