Detect source type (HTML / XML)

[Kroc]

In order to avoid XML being broken by the HTML filtering, we need to detect (or provide via parameter) whether the input text is HTML or XML. We can check for a DOCTYPE vs. an XML prolog.

Note that this also needs to be done for the inserting HTML function that appends a DOM fragment. This is more tricky because the fragment won't include a DOCTYPE / prolog. -- we probably need to remember the original type detected and re-use it there.

This can be expanded in the future to detect between XHTML / HTML4 / HTML5 & XML for more specific filtering (entities, UTF-8 &c.)