Unfortunately, we cannot yet parse the additional hints (plural forms,
cases, parts of speech, etc., everything inside <repr> tags) because the
XML::Simple API does not retain the correct ordering of CDATA content
mixed with subtags... :-/
This also removes the options -SpellTolerance, -Morphology, and
-CharTolerance, which are no longer supported by the new API (as long as
I can see.)
Note that although the XML contains UTF-8 data, XML::Simple decodes
it to latin1, so we have to re-encode it to get good results.
