Assess behavior of BreadthFirstSearch on mixed directed/undirected graphs

[nicolaspayette]

Having to do 11b7434 to preserve behavior after fc7a997 while working on #91 made me wonder if the current behavior was correct.

Presently, unless a graph context is directed (i.e., unless the context is set to use a directed link breed), only undirected links are followed by BFS. We probably want to be smarter about that and follow undirected links and links in the right direction.

[nicolaspayette]

65b1a0e makes unweighted path finding (path-to, turtles-on-path-to) follow both undirected links and directed out links. A user wanting to follow only undirected links can set the context accordingly.

2a3d061 makes the behavior of turtles-in-radius coherent with path finding primitives. I have added a turtles-in-undirected-radius to get the old behavior, but I am not entirely sure it was the right thing to do, since this can be achieved with the proper context. Maybe we should just have turtles-in-radius and turtles-in-reverse-radius and handle everything else with set-context. I think it would be simpler overall.