allow specifying symbolic stop lane

[namdre]

either via lane="EDGE_first" (ugly)
or via edge="EDGE" lane="first" (also ugly)
or something else. see #5115

[namdre]

I think the second version is ok.
As long as the symbolic names have no underscore, we also don't run the risk of name collision between symbolic names and lane ids. In contrast, to #7584 we can always interpret the stop lane in the context of a know vehicle so "first" can always be interpreted in regard to a vClass. (similar to departLane).

The advantage of allowing stop attribute 'edge' (with lane then optional) is also a much cleaner definition of person stops.

[namdre]

added tests in [5e773e5]

fixed in [781b057]

setting attribute edge instead of lane now uses the rightmost allowed lane.