* Don't refresh eth_dst rules if a non-hairpinned host didn't move ports (reduced maintenance flowmods)
* Implement Generic hardware type (relative table IDs, no TFM, barriers)
* Decreased L2 cache updates (reduced maintenance flowmods)
* Controller flood requests use same packet with multiple output ports (reduced packet outs for nexthop resolution)
* Avoid unnecessary regeneration of Prometheus learned_macs.
* Add Prometheus variable faucet_service_secs for background services runtime.
* Upgrade docker base image to alpine 3.8.1