Includes new add method that uses .clone() for support.

[niftynif]

I implemented an add method for the btree in progress. It is intended to be refactored later using an alternative to .clone() that passes the borrow checker, but for now, it works as intended. r? @catamorphism

[huonw]

You do realise that this is cloning the whole tree, right? i.e. every insert is requiring duplicating everything, which completely destroys the O(log n) behaviour. I am personally very against merging something like this; would it be possible to do the work to avoid the clones now, before merging?

Also, almost all of these methods should be taking &self or &mut self to modify in place rather than consuming the tree, which will probably make it much easier to avoid .clones. (Compare to the various methods of extra::treemap, e.g. .insert.)

E.g. I'd expect signatures like:

impl<K: TotalOrd, V> BTree<K,V> {
    pub fn find<'a>(&'a self, k: &K) -> Option<&'a V> { ... }
    pub fn insert(&mut self, k: K, v: V) { ... }
}

(See the Map and MutableMap traits for the signatures & method names that are conventionally used.)

[niftynif]

I understand where you're coming from, @huonw, and I'd like to add that this work you see here is not meant to represent a finished product. This would of course be revised extensively to hopefully get rid of every instance of .clone() in the insert method, but the goal of this PR is to add new functionality that hadn't yet existed for this data structure, and to get the algorithm for insertion "on paper." I originally began the implementation of the insert method with &mut self in the method signature, but this was unsuccessful in my first few attempts and @catamorphism suggested that the method should take self, rather than a reference to self.
I understand if you're not convinced by that explanation. Maybe @catamorphism can do a better job of explaining the goals and process behind this PR.

[brson]

Needs rebase.