Refactor element/text base class

[teleological]

Factored out common methods in Element/Text. Couldn't come up with a decent name for the result; would welcome suggestions.

Perhaps more importantly: Noticed some memory issues when adding text siblings. It may be good to patch those before letting this out into the wild. Working on a fix, but wanted to get a clean refactor out first. ( Bug fixes in #362 )

[teleological]

Renamed "fraternal" nodes to "non-attribute nodes", updated XmlComment to be a subclass, and changed XmlNode::New so that it creates the correct kind of subclasses (fixes #226)

[rchipka]

It could also be called a proto-element or ProtoElement type. Also, in #360 the add_child, add_next_sibling, add_prev_sibling, and import_element functions have all been moved to XmlNode. The reason I moved these to XmlNode is so that the refcount logic is all in one location. This way, anything that inherits from XmlNode will have these functions available, but won't have them exposed via the API unless the XmlNode inheritor explicitly implements addChild(), etc.

We could easily move that same logic into this sub class instead, just as long as it's in one place.

[teleological]

@rc0x03 I also moved import_element to XmlNode, but didn't move the others because they didn't seem a good fit with XmlAttribute, which also descends from node. Have you looked into that? I've started to review #360.

[teleological]

I'm going to extract the memory fix for adding/merging text nodes into a separate PR that can be merged without delay. I will leave the refactoring of the Element/Text base class for later consideration. The correction of the wrapper selection for child nodes, etc. is technically a bug fix, but it also impacts usage, since nodes returned from some operations won't respond to the same API as before the change. So it would be kinder not to shoehorn it into a minor release.

[teleological]

Bug fixes extracted to #362. Once that is merged, will rebase this PR to exclude those commits.