Implements min_number for segmentation

[runarberg]

This is implemented with a simple algorithm:

• Compute all distances between subsequent vertices (segments) in the geometry.
• While the number of vertices is less then the desired number:
  • Find the longest (max₁) and the second longest (max₂) segments.
  • Find lowest number (n > 1) such that max₁ can be split into evenly into n segments which are shorter then max₂.
  • Split max₁ into n equal segments.
  • Add the new vertices to the geometry.
  • Update the distance array with the new distances.

resolves #405

[runarberg]

I change the algorithm a little from what I originally proposed in #405. The benefits of this one is it is a little more optimized (will require fewer iterations) and it will add vertices in very similar places as max_distance (I think they start to differ when min_number revisits a segment it has split before).

I also tried to make sure to retain the inner structure of the geometry (e.g. if using Float32). Other methods don’t do that, but I felt it was worth it to try it. I am using Float32 in my model but simplify will change my vertices to Float64. If you are fine doing it this way I may create a separate issue to this in more places, including simplify.