Use iterator (not index-based) code for streaming document output

[rolfl]

SOme legacy JDOM 1.x applications took advantage of the fact that the XMLOutputter used an iterator-based mechanism for processing Element content. JDOM 2.x swapped to using index-based lookups.

This swap is 'reasonable', but, using iterators makes sense too, if it works.

See discussion here: http://markmail.org/message/ejxecyfabnrccxka