BGP: LLGR Staleness optimization dropped.

This brought unnecessary complexity into the decision procedures while the
performance aspects weren't worth it. It just saved one ea_list traversal
when many others are also done.