Refactor module structure

[PoignardAzur]

This echoes a discussion I had with Chad a few weeks ago. Parley's current structure de-facto splits into three types of items:

• input data structures.
• output data structures.
• algorithms mapping the former to the latter.

(EDIT: There's also data structures for text editing which are mostly separate from the rest.)

This PR makes these four blocs into explicit modules, which should make both the code and the documentation much easier to read.

This should also help people navigate the upcoming refactor; algorithms will be changed radically, input types may be changed, output types are likely to stay the same. Explicitly marking them should make parallel work easier.

Another thing the PR does in de-tangle layout types. Currently, when you're looking for Cluster, do you find it in cluster.rs? Nope, layout/mod.rs. What about ClusterData? layout/data.rs. Splitting these types into their eponymous files makes them easier to find without IDE assistance.

[PoignardAzur]

I think this is ready to merge. I'd recommend reviewing the commits individually otherwise this is a lot to take in at once.

[PoignardAzur]

To be clear, this is a first stab at sorting parley's files differently.

Follow-up PRs should:

• Move a lot of code from outputs/ to algos/.
• Flatten the directory structure a bit.
• Document all of this, including an ARCHITECTURE.md file.

[waywardmonkeys]

I think this is blocked (at least) pending a comment from @dfrg …