Feature: Concurrency under-the-hood explanation in Docs

**Describe the solution you'd like**

It would be great for users of the library if you could explain prefetch vs workers in the docs for controlling the concurrency of these brokers. In particular, if n prefetch means n processed at once etc. or just n 'held locally' by the consumer

Knowing the under-the-hood working can help us in fine-tuning the brokers

Many thanks