Iterweb

* Python 3.8 changes how functools.partial() wraps async functions, needed
a minor patch

* read response.body instead of response.text to allow downloading of binary files
* track_urls is now True by default

* lots of refactoring
* now enqueue Requests instead of urls
* Request can now hold a callback, useful for Requests emitted by parse()
* callback can be a coro instead of only an async generator
* Spider can now track urls so the same page isn't hit twice
* page spidering is more concurrent, results are still in order
* all tests use mock requests, found and filed a bug in aiohttp (4684)
* client_factory is now just a client object again

* use a Queue to hold urls
* move pipeline functionality into its own class
* use a client_factory to create ClientSession

* did I really call it DropItemBad? DropItemError

* adding DropItemBad, change logging levels