This patch prevents accidental leakage of content-type header and charset
between calls to get_title. Without clearing these, it is possible for a URL
title to be decoded from the wrong charset if a URL was previously titled
with a differing charset to the current one. This patch clears these stale
values to guarantee accurate charset decoding per URL.

This patch adds the ability for URL_Title to fall back on the Content-Type
meta http-equiv tag, or failing that, the Content-Type HTTP header itself.
This should improve correctnes when dealing with HTML documents other than
HTML5.

This patch moves the HTML entity decoding until after the raw bytes from the
HTML document are translated through charsets. Previously, entities were used
as decoded by the HTML parser into UTF-8, which meant that non-UTF-8-encoded
strings from documents could become mixed with UTF-8 characters, making
the subsequent character encoding transformation impossible to perform
correctly.