Refactor Namespaces class to use a stack-ish structure

I had begun working on this issue and currently I removed the previouse implementation that create it's own list a like collection of namespaces within single Namespaces class. Hovewer, as I understand this is not the actuall thing the was on Your mind?

So how would You like to remove the _parent references and replace them with some other stack structure?

Note: the attached patch already uses the Assert class attached here: CASTOR-3138.

Should I introduce something like a 'NamespacesStack', that will have similar methods to the Namespaces class, but would rather traverse the entire stack of namespaces?

I added the seperate class for organizing the namespaces into stack structure, I didn't remove the _parent field from the Namespaces class, due to the fact that there are still classses, apart from NamespaceHandling, that relay on this structure.

Yes, that's very much the implementation of what I had in mind when creating this issue as a reminder a few weeks ago.

Having said that, it looks like it's not that simple due to rather odd code usage. Can you please analyze what classes are using this structure ?

And one more observation from my past 'playing' around with Namepaces. I noticed that new Namespaces instances git created (and linked into the 'tree') even when there was no namespace declaration on the element being processed currently - just to be removed with the corresponding ending element. This, of course, can cause a big overhead wrt to object instantiation and garbage collection. So maybe one can be a bit smarter here.

I had found fallowing usage of Namespaces class:

• Marshaller#marshal(Object object, XMLFieldDescriptor descriptor, ContentHandler handler, final MarshalState mstate) line 1309

• AnyNode2SAX#processAnyNode(final AnyNode node, final DocumentHandler handler) line 168

• AnyNode2SAX2#processAnyNode(final AnyNode node, final ContentHandler handler) line 242

• DocumentHandlerAdapter#startElement(String uri, String localName, String qName, Attributes atts) line 185
  and DocumentHandlerAdapter#startPrefixMapping(String prefix, String uri) line 208

• SAX2ANY#startElement(String name, AttributeList atts) lines 197 and 257
  SAX2ANY#startPrefixMapping(String prefix, String uri) line 173

and finally

• Sax2ComponentReader#startElement(String name, AttributeList atts) line 239

It finally works!

Please double check if all the tests run correctly and I didn't by accident break something.

Jakub, could you please attach a new patch, it applying the original one is nearly impossible (with too many things having changed).