Explore parallelization

[GabeDottl]

Pretty complex due to dependency graph. Also made messy by Python's concurrency model.

Could perhaps do a minimal version w/async + module loading?

E.g. determine all module dependencies, create a producer of modules which processes modules as soon as all dependencies are met.

Could become a memory nightmare left unchecked - would be good to add in some restrictions to prevent consuming all CPU or memory..

Or perhaps only run this way when we care about maximal throughput? (e.g. on servers)

[GabeDottl]

Also interesting to consider whether there's a way we could make it so all modules essentially became independent of each other. The greater degree to which we can rely on stubs + lazy modules, the more true this is..