Pathfinding: add support for goals made up of multiple nodes

[davebaol]

See discussion and proposed implementation here, #7 (comment)

Ideally, this kind of goal could be specified by an explicit set of nodes or abstract properties.

[piotr-j]

Any progress on this issue? Would be quite useful.

[davebaol]

@piotr-j
No, sorry. Rather busy these days. Hopefully I'll have the time in the near future.
PR welcome, of course. 😄

[piotr-j]

Ill look into it once I figure out simple working ai, perhaps next week.

[piotr-j]

Does arbitrary list of endNodes makes sense? Since there is no relation between, they can be on opposite sides of the graph and that would result in basically endNodes.size searches that could as well be separate searches. Unless only connected nodes ware valid or something like that.

As abstract properties, something like any node within maximum cost from endNode would make some sense. That would probably require full search and then backtracking till we find node that exceeds the specified cost.

Any thoughts on that?