sampling-tree 0.1.0

A simple sampling tree implementation for sampling discrete distributions with sparse dynamic updates. This allows us to sample efficiently from a distribution given the relative importance of each datapoint. Construction time is O(n), updating is O(log(n)), and sampling is O(log(n)). The memory footprint is no more than twice the size of `n*std::mem::size_of::<T>()` where `T` is weight datatype.