Feature: worker pooling

Hello,

It would be great if I could create a pool of workers. That way, parallelizing hot paths could be completely abstracted using `workerize`.

Considerations:
* To implement worker pooling, a worker may have to know when another worker is busy, so it can handle take over handling new requests until becomes busy itself
* Since workers independently manage memory, they would all have to run the same code, and there would have to be a way to run a function on all workers simultaneously in order to share state

It will be interesting to see what we will need to do for this. Would you accept a PR?

Thanks for considering!

The issue names no files, tests, or entry points, so begin by locating workerize's worker creation and request-dispatch code. Define how workers report busy state, take over requests, and run a function across all workers; done requires a documented, working worker pool with those semantics.

索引モデルが issue の本文から書いたものです。