delaunay.find and coincident points

[Fil]

fixes #55

[mourner]

I wonder if there's a deterministic version of the fix...

[Fil]

I thought of a list of visited nodes (to avoid those if we get to another dead-end), but then it costs memory… and too much code.

[mourner]

Not a perfect approach too, but perhaps a custom seeded random would work?

[Fil]

I think (i + 1) % length would be correct, in fact, in all cases… I had started with that, then I realised that I had to make 100% sure it would never fail.
But, there's no way the step ever goes from a normal point to a coincident point. So this call happens only when the initial hint i is coincident, and the only way it could loop forever, if we enumerate points with i => i+1, is if we had only coincident points — an impossibility. Right?

We can make doubly sure by memorizing the initial i and preventing any loop which could happen in the degenerate case "no valid point".

[Fil]

closing this branch in favor of a local d3 branch