immerrr again

@raelgc there's a rather active backport of webkit that supports fresh qt at https://github.com/annulen/webkit, but I have never tried to build it

FTR https://github.com/GoogleChrome/puppeteer/issues/1325 was closed in June

Another point to consider: what if there's a redirection loop? Limit the redirection chain? Also, how to report this to the user?

Here's a simple example to reproduce. A bit contrived as it plays around query parameter ordering, but does the job: a browser can open the page, and the spider cannot...

@lborowczak the issue is not about `canonicalize_url`, it could be a certain site that redirects you to put some session cookie to a different page, and then gets redirected back,...

@kmike you beat me to it. again :)

@kmike what if the scheduler consults `request.meta['redirect_urls']` and doesn't filter requests that have `request.url in request.meta['redirect_urls'][:-1]`? It will kinda introduce this implicit requirement on the functionality of redirect middlewares, that...

Another slightly subtle application of the context node being current node by default: ``` sel.xpath('//div[rel-id("some-id")]') ``` which means select a div, that contains an element with `id="some-id"`

I have implemented the gunicorn logging in a very naive and straightforward way in [my branch](https://github.com/encode/uvicorn/compare/master...immerrr:implement-gunicorn-access-log-format). Would you be interested in a PR?

@euri10 sure, np, here you go: https://github.com/encode/uvicorn/pull/947