Comprehensive explanation of `download_dataset`

I suggest writing a step-by-step documentation on what is going on behind `download_dataset`. The source code is very hard to follow (I am blaming myself). We had multiple discussions about the process already and that in itself suggests that it is anything but transparent.

I think there will be 2 benefits
- Outsiders will be able to understand the process
- Helpful for developers to discover bottlenecks and bugs and make it easier to improve the process