Consider changing add_edge(src, dst, e) to add_edge(e: impl IntoEdge)

[pnevyk]

Advantages:

• Cleaner (?) API for unweighted graphs: graph.add_edge((src, dst)) instead of graph.add_edge(src, dst, ()).

Disadvantages:

• Worse documentation? add_edge(&mut self, src: VertexIndex, dst: VertexIndex, edge: E) is much clearer than add_edge<T: IntoEdge<VertexIndex, E, Ty>>(&mut self, edge: T).
• Potential problems with type inference. Especially in case of encapsulations where we don't use G::VertexIndex directly, but VI: Borrow<G::VertexIndex> instead.

---

In summary, I am not much in favor of this change, but just wanted to leave a note if I change the opinion and the disadvantages will turn out not that bad.