This needs a title

Highest priority:

See automata_scratch/examples.py and implement some of the tougher examples.
- spiral_nested_2 & spiral_nested_3 (how to compose efficiently?)
- twisty_torus
- ram_horn_branch - Can I pass depth via a closure? Do I need mutual recursion for this?

Elegance & succinctness (my recent closure work may help with this):
- Why must I repeat myself so much in these definitions?
- What patterns can I factor out? I do some things regularly, like: the clockwise boundaries, the zigzag connections, the iterating over a Vec<Vertex> to transform each element and make another vector.
Docs on modules
Grep for all TODOs in code, really.
Look at performance. Can I save on copies of geometry by using Rc<OpenMesh> or the like? In many cases I have nothing but copied geometry. Can I pre-allocate vectors instead of extending/appending?
- connect() is a big performance hot-spot: 85% of total time in one test, around 51% in extend(), 33% in clone(). It seems like I should be able to share geometry with the Rc (like noted above), defer copying until actually needed, and pre-allocate the vector to its size (which should be easy to compute).
Look at everything in README.md in automata_scratch.
I can't really do mutual recursion with the closure method, can I? I'd need actual functions for that.
N.B. "Constants" outside the closure only work the way I think they should work if:
- they're actually static
- they implement Copy
- the closure can move them

Fix links in tri_mesh docs that use relative paths & do a PR?
Look in https://www.nalgebra.org/quick_reference/# for "pour obtain". Can I fix this somehow? Looks like a French-ism that made its way in.
Multithread! This looks very task-parallel anywhere that I branch.