Here are two implementations of a method to calculate the width of a number in bits,
sort of the inverse of calculating 2 ** n. There were eight submissions. I have
retained the slowest a method that uses Math.log, and the fastest (19 times faster)
that uses a fully unraveled binary search.