Reimplement YinStatementSourceImpl in terms of DOM

Current implementation of YIN parsing is based on StAX, which means that each processing phase ends up re-parsing the XML from scratch. It also means that YinStatementParserImpl is very complicated.

Using a W3C DOM Document is much more natural, as that will end up parsing the input XML only once and provides convenient iteration, leading to a much simpler and performant implementation at the cost of some memory overhead.