HTML character encoding in the context of HTTP is quite tricky to get
right and honestly, I'm not sure if I did get it right this time. But I
think, the current behaviour matches best what web browsers try to do:
1. Let Requests figure out the content from the HTTP header. This
overrides everything else. We need to "trick" LXML to accept our
decision if the document contains an XML declaration which might
disagree with the HTTP header.
2. If the HTTP headers don't specify any encoding, let LXML guess the
encoding and be done with it.
8768ff07b6