Always return a cycle in digraph_find_cycle if no node is specified and a cycle exists

[IvanIsCoding]

Closes #1171

Instead of picking an arbitrary node if it is not provided, we find a smarter pick that is in a strongly-connected component with more than one node. It also handles the case of self-loops in case there is a SCC with a single node.

[coveralls]

Pull Request Test Coverage Report for Build 9103568887

Details

• 25 of 25 (100.0%) changed or added relevant lines in 1 file are covered.
• 6 unchanged lines in 1 file lost coverage.
• Overall coverage decreased (-0.01%) to 96.475%

---

Files with Coverage Reduction	New Missed Lines	%
src/shortest_path/all_pairs_bellman_ford.rs	6	95.53%

Totals
Change from base Build 9103534121:	-0.01%
Covered Lines:	16748
Relevant Lines:	17360

---

💛 - Coveralls

[mtreinish]

Overall I like the idea here it makes sense to me to use a connected component function to get a node in a cycle and go from there. I just have some concerns about the backwards compatibility implications. They're likely unavoidable to use the scc functions, but we should document it as an upgrade change then.

[IvanIsCoding]

Overall I like the idea here it makes sense to me to use a connected component function to get a node in a cycle and go from there. I just have some concerns about the backwards compatibility implications. They're likely unavoidable to use the scc functions, but we should document it as an upgrade change then.

I removed most of the traits but petgraph::visit::Visitable is a requirement for kosaraju_scc and by https://docs.rs/petgraph/latest/petgraph/visit/trait.Visitable.html most graphs implement that

[mtreinish]

LGTM now, thanks for the quick update.