Improve the deserialization support by making the extension point consistent

[drmohundro]

It is difficult to know when to implement XMLElementDeserializable versus XMLIndexerDeserializable. From the perspective of someone implementing custom deserialization, the indexer is for nodes that have child elements (i.e. a class with properties). Element deserialization is almost always for individual nodes (i.e. strings, ints, dates, etc.).

It'd be great to make these consistent. This has also been brought up as a point of difficulty and confusion in the library itself.

See https://stackoverflow.com/questions/45851987/deserialisation-error-with-swxmlhash as an example of the confusion.

(see also #84)

[drmohundro]

See PR up at #259. It should resolve this. I've renamed XMLIndexerDeserializable to be XMLObjectDeserialization - its purpose is to deserialize object structures like classes or structs. So, multiple values. Then there is a new protocol named XMLValueDeserialization - it serves the purpose of XMLElementDeserializable and XMLAttributeDeserializable. It is only used for custom value deserialization. I hope this will clarify the usage a lot. It makes sense to me anyway!