Handle missing title from table of contents nodes that have children

[samuelclay]

Using lxml's .text_content() instead of text because some ToC nodes look like this:

<li>
    <a href="text/chapter-1.xhtml"><span epub:type="z3998:roman">I</span>: Looking-Glass House</a>
</li>

This PR fixes that issue by using lxml's nested text extractor instead of using the root node's text alone.

[aerkalov]

Thanks for this!