added support for weighted average_path_length

[FdeFabricio]

The same way we can provide weights for the diameter, it would be a good idea to have a weighted average_path_length

[ntamas]

The underlying C library does not provide a weighted average path length function, only an unweighted one. (The diameter function exists in both variants: igraph_diameter() and igraph_diameter_dijkstra()). The only way to get around it at the moment would be to calculate the full path length matrix and then do the averaging in the Python layer.

[szhorvat]

@ntamas This is now implemented in the 0.9-series C core as igraph_average_path_length_dijkstra(), but it's not exposed in Python yet for some reason.

Also, the current docs are now wrong when they say,

Otherwise for all unconnected vertex pairs, a path length equal to the number of vertices is used.

Adding to 0.9.9 milestone, if that's okay.

[ntamas]

Implemented in 562c05f