Add vertices_at_distance and next_nearest_neighbors

[JoeyT1994]

PR which moves the nth_nearest_neighbors functionality from ITensorNetworks.jl to the GraphsExtensions module in this library.
A few short tests are added, although maybe more of the GraphsExtensions code should be tested in the future (leaving for another PR).

[mtfishman]

We should assess the name of this function.

Apparently the terminology "k nearest neighbors" has a different meaning in the graph community. For example, see the function k_nearest_neighbors in older versions of NetworkX, which was renamed to average_degree_connectivity in NetworkX v3.

The closest function in NetworkX that I can find to the functionality we want here is descendants_at_distance, though the terminology "descendants" makes it seem like it is for directed graphs.

Perhaps we could call it vertices_at_distance. Additionally, it may be better to implement it using Graphs.neighborhood_dists. In that way only a single neighborhood needs to be computed, and then we can filter the vertices at a certain distance, possibly using SplitApplyCombine.jl.

[JoeyT1994]

vertices_at_distance seems like a reasonable name to me. I guess k_nearest_neighbors is using k to mean degree but that is not a very transparent name.

Okay yeah it should be easier to write using neighborhood_dists rather than recursively calling the function. Let me change that.

[mtfishman]

Thanks!

[mtfishman]

Thanks!