Since we keep track of the size of subnodes, `at` should be pretty easy to achieve `log(n)` indexed lookup.