Drop dependency on readtime

[lemon24]

... and implement/vendor it directly in the plugin.

Why?

• readtime 2.0 requires pyquery requires lxml, and that isn't exactly available everywhere (easily).
• The logic is trivial, 20 lines, with another 20 to convert HTML to text and count images.
• We already use BeautifulSoup to convert HTML to text (and parse HTML).

[lemon24]

To do:

• [x] BeautifulSoup implementation
• [x] extract common stuff to reader._html_utils
• [x] update dependencies: setup.cfg, run.sh, tox.ini
• [x] docs
• [x] changelog