fix bug with flipEdgeNetwork created by piecewise Dijkstra

[Joel-Hecht]

Fixes #209

For some vertices A,B,C using shortestEdgePath() to connect vertex A to vertex B, and vertex B to vertex C would result in paths AB and BC that possibly contain one or more duplicate edges (i.e. one or more edges in AB has its twin in BC). The flipout algorithm is not able to flip these 'back-and-forth' loops (or the edges immediately before/after them), leading to incorrect shortening of the input path.

PR adds an option (enabled as default) in constructFromPiecewiseDijkstraPath() to remove any subset of the resulting curve that starts and ends at the same vertex, which, for all halfedges in the subset, also contains their twins.

Only fixes the 'artifacts' problem in the linked issue. Other 'problems' mentioned in the linked issue are fundamental misunderstandings of the algorithm.

Also adds a few contrived tests illustrating cases that would fail before, but now have the expected behavior. These tests are not very interesting and can probably get deleted.

Possibly also fixes this issue noted in comment in wedgeIsClear() (?):
TODO handle checks in case where the path lollipops out and back along a single edge

[nmwsharp]

Thank you for this! The problem with the "out-and-back" piecewise Dijkstra paths makes sense. I see how it could happen and agree this simple change helps a lot at least.

We can't apply it when marking vertices, but otherwise I think it should always be enabled. I pushed a commit to this PR that makes that tweak, as well as some other cleanup around the code style and test organization, but the algorithm should be unchanged.

I'm going to go ahead and merge, let me know if you spot any issues I introduced in translation.