PyPump assumes URLs

[tsyesika]

For the most part assuming the URLs is probably OK however, because pump provides the links in the link section of some API calls and also via some webfinger discovery endpoints, we probably should attempt to use those first if we can.

[tsyesika]

I think that there should be a object which produces URLs from wherever they might be. The initial stuff for example the client registration URL should be found at https://microca.st/.well-known/host-meta.json I could then just do URLs.client_registration.

This may actually fix #62 because class' wouldn't have ENDPOINT variables they would be found on this URL object with fairly descriptive names.