`KDNode.dist` returns squared distance while `search_nn_dist` requires euclidean distance

The two lines below, one requires dist to be euclidean distance, and the other requires it to be squared distance:
https://github.com/stefankoegl/kdtree/blob/master/kdtree.py#L501
https://github.com/stefankoegl/kdtree/blob/master/kdtree.py#L508

And the `best` argument is not used
https://github.com/stefankoegl/kdtree/blob/master/kdtree.py#L517