LeveledSpanningTree

[dellaert]

New data structure that supports efficient computation of fundamental cycles by:

1. Constructing a BFS tree from a root node
2. Recording both parent and level information for each node
3. Using level information to quickly compute paths to the least common ancestor

This follows Paton's 1969 algorithm for finding fundamental cycles in undirected graphs,
in "An algorithm for finding a fundamental set of cycles for an undirected linear graph".
However, the level information allows O(V) computation of fundamental cycles where V is the
number of vertices in the graph. Without level information, finding the lowest common
ancestor would require O(V²) time in the worst case.

Added three tests for three different ways to call (adjacency, edge list, factor graph).