Simpler power(_:) function?

[oscbyspro]

After some experiments in (#71), the difference between arbitrary integer a.power(b).times(c) and a.power(b, coefficient: c) appears small when the initial power is without heap allocation. The former is simpler to test and reason about, so I prefer it. There are at least two ways to prevent this heap allocation: (A) adding a static protocol requirement or (B) waiting for a small-integer optimization (#44) that puts it on the stack.

[oscbyspro]

Some values may benefit from static protocol requirements. 0, 1, and 10 are the first that come to mind. But I'm hoping that a small-integer optimization (#44) is the answer because it is a more general solution.

[oscbyspro]

Looking at the assembly output of a similar Swift.UInt algorithm, it looks like the compiler turns the "coefficient" version into the "non-coefficient" version by setting the initial power to 1 and inserting an imul before the loop. As a result, the coefficient version contains one more instruction. Silly.

Edit: But this is only in -O mode. In -Osize they are basically identical, which means that we do get one multiplication for free in the "coefficient" version. Edit (again): Never mind, it isn't free because we still have to put the coefficient on the stack to call the function :(