HtmlCleaner shouldn't transform content when creating a Document

When an HTML is parsed and serialized in a w3c DOM Document, some attribute values are transformed in XWikiDOMSerializer: for example an image source like http://foo?a=b&b=a will be transformed to http://foo?a=b&b=a.

This kind of transformation shouldn't be done at that stage since the Document should reflect the exact content of the page: this kind of transformation - if needed - should be done only when serializing the Document back to HTML.