Multiple implementations of ietf-lisp-address-types cause bundle refreshes

This is a sub-part of HONEYVBD-14 but this has higher severity as this can lead to bundle refreshes, which is what makes SingleFeatureTest fail in Nitrogen.

The current issue is that not only VBD currently hosts the same Yang model which has been already added to mdsal.models, but it uses a different Java implementation of SimpleAddress union builder. Karaf4 detects that and attempts to wire the two bundles together. See [0] for more details.

Only one of the implementations should be chosen, and hosted in Mdsal.

[0] https://bugs.opendaylight.org/show_bug.cgi?id=8923#c13