Support for DOM and other external tree models

Support for the namespace axis has been reinstated in JDOM. This underpins functions such as get-in-scope-prefixes, and ensures that namespaces are properly copied by xsl:copy-of. {axes-jdom049, 055, 129 etc}

For ease of testing, a new command line interface net.sf.saxon.jdom.JDOMTransform has been added. The arguments are exactly the same as the normal net.sf.saxon.Transform command.

The constructor for class jdom.NodeWrapper has been made protected. A new wrap() method has been supplied on the jdom.DocumentWrapper class, allowing any node in the document to be wrapped, provided that the document node has been wrapped.

Whitespace stripping should now work for JDOM input in the same way as for other tree models (see below).

A set of DOM wrapper classes have been written, analogous to the JDOM wrapper classes. {axes-dom[001-nnn]}

The DOM wrapper has been tested with the Crimson DOM provided in JDK 1.4 and with Xerces 2.5.0. Different DOM implementations are known to vary widely. Saxon's DOM interface does not attempt to deal with entity reference nodes, which appears to be OK with the default configuration of these two parsers. CDATA sections are treated as text nodes, no attempt is made to merge them with adjacent text nodes.

The DOM interface is very inefficient, for example it has to resolve namespace prefixes by searching the namespace declarations every time a node is referenced. Don't use it if performance matters to you.

I have changed the code for writing output to a DOMResult so that it now uses DOM level 2 interfaces Document#createElementNS() and Element#setAttributeNS(). As a result, it should now be possible to use methods such as getPrefix and getNamespaceURI on these nodes.