[url] Don't try and build an etree out of non-html

[Stefano Rivera]

Some digging through memory logs shows huge spikes of "Element" objects causing memory to be allocated.

These correlate strongly to the presence of .jpg URLs in the logs.

Probable culprit:
    def _get_title(self, url):
        "Gets the title of a page"
        try:
            headers = {'User-Agent': 'Mozilla/5.0'}
            etree = get_html_parse_tree(url, None, headers, 'etree')
            title = etree.findtext('head/title')
            return title