We now plot the cell many contours as a single patch object relying
on the special low-level syntax of patch.

We want to trim the fat.

The goal is to be able to plot the tissue with one face per cell, to
have a simplified representation of the cells on the tissue with e.g.:
>> patch( 'Faces', F, 'Vertices', V, 'FaceColor', 'g' )

Brute force, thanks to the power of linear scattered interpolant.

We need to check that an argument is not empty before checking whether
it is nan.

The original mesh file coming from MorphoGraphX had dozens of millions
of vertices and faces and was not aligned with the segmentation data.
So I downloaded MeshLab and simplified and recentered the mesh.

WARNING: This is a simplistic reader that can only handle one kind of
data formatting. The file must be a binary encoded following the
binary_little_endian format, there must be 3 floats per vertex, and 3
integer per face, nothing else. Any variation will result in a crash at
best.

Everything else I could find was super slow.

Extra speedup factor of 2.

One call to pdist per run. Speed-up ~ x8.

That can connect unordered points on a non-convex polygon,
provided that connected points are close.

Created by [gh-md-toc](https://github.com/ekalinin/github-markdown-toc)

Created by [gh-md-toc](https://github.com/ekalinin/github-markdown-toc)

Created by [gh-md-toc](https://github.com/ekalinin/github-markdown-toc)

I simply copied the part from the README.md.