Split the range of a node into the ranges of the tokens that delimit it, so the transformer can give an opening or closing token a `SingleTextNode` of its own. `size` is the length of the token in characters: 1 for `(`, 2 for `[|` or `{|`. The ranges are computed from the ends of the node's range, not looked up, so they are only right when the token really is the first or last thing in the node.