Implementing iterative removal of non_terminal_leaves in Steiner Tree approximation

[OrionSehn-personal]

This should resolve #6881

Might not be the fastest possible solution, it checks the neighbors of non-terminal leaves, and removes them in bunches of sets using set comprehension and differences.

I added a relatively simple test to ensure that this function acts as intended, and trims down any non-terminal leaves, but any consequential non-terminal leaves created by that trimming.

This might slow down the approximation significantly however, I didn't do speed testing on this, and I'm not sure if we should consider adding an optional flag in the Steiner tree approximation to perform this additional optimization.

[rossbar]

@OrionSehn there is already a PR open for this issue at #7136 with a fair bit of discussion - perhaps you could take a look at that one and weigh in there?

[rossbar]

This approach makes sense to me and the test adequately demonstrates the problem with the current implementation (i.e. non-terminal nodes along a path become terminal with subsequent removals).

There are potential "gotchas" related to this approach implicitly assuming that leaf nodes have degree one. This may not always be true: two cases I can think of are 1) graphs with self-loop edges (see example) and 2) MultGraphs with multi-edges.

# Desired behavior
>>> G = nx.path_graph(10)
>>> _remove_nonterminal_nodes(G, [4, 5, 6])
>>> list(G)
[4, 5, 6]
>>> H = nx.path_graph(10)
>>> H.add_edge(9, 9)  # self-loop on one terminal node
>>> _remove_nonterminal_edges(G, [4, 5, 6])
>>> list(G)
[4, 5, 6, 7, 8, 9]

Since this function is only to be used in the context of the steiner_tree approximations, these may be fine - but it's worth double-checking how steiner_tree handles these cases to make sure the _remove_nonterminal_leaves gives consistent results for these cases!