PyWebXML documents could be more lazy

[horazont]

Often, we need only the XML representation of a PyWebXML document. It would be really smart to create a Document subclass to be returned by the PyWebXML parser which only actually parses the XML if some of the semantic attributes are requested (which only the blog actually does if I recall correctly).

This would save a lot of parse-unparse performance, especially as the intermediate product of any non-raw XSLT done in PyXWF is held as a PyWebXML document.

[horazont]

This would require to finally define the semantics of copying. We would have to make clear which operations require a tree for them alone (that are all operations doing in-place transformations or directly reusing elements from Document attributes) to avoid glitches.