adding triangulation and tetrahedralization functions

[odedstein]

This PR adds triangulation and tetrahedralization capacity.

For triangulation, this uses CDT. This is, in my opinion, better than Shewchuck's classic Triangle for 3 reasons:

• It has MPL2 licensing, like libigl, and unlike Triangle, which has a custom license that is very restrictive.
• It auto-detects holes, which you need to specify seed points to Triangle for.
• It is very robust.

For tetrahedralization, this uses TetGen through its libigl wrapper.

It's still not quite ready to merge before we discuss 2 important points:

• CDT does not yet support refinement (needed for min angle and max volume options). They seem to be actively working on it, and the library is so nice that I suggest we stick with it anyways. This is the active PR. I tried it, and it works fine, but I do not want to use a non-merged PR for our library.
• TetGen is not just GPL (our normal copyleft), but AGPL, which adds an additional, more restrictive license to our copyleft module. I would have liked to used the new paper CDT (this has the same name as the 2D triangulation library...) for this, which seems to be a more modern TetGen, but their code does not support refinement of tet meshes, which is needed for quality and maximum volume options.

[sgsellan]

There seems to already exist a great python wrapper for tetgen, with a good API that is similar to ours. So I personally do not think the damage of polluting the license is offset by any gain of having it wrapped in our library, at least right now. So I propose we excise the tetrahedralization from this PR.

On triangulation, I am happy to learn about this library! Seems to do what I use triangle for, but better. Happy to both merge now, and merge once their refinement PR is merged. Up to you.

[odedstein]

Your point makes sense to me. Let's leave off this pull request for now, and revisit it when the triangulation library is more mature.