Should frobenius_map powers be unrolled

[ValarDragon]

Currently we have a method fe.frobenius_map(power), which computes a given power of a frobenius map. This involves multiplying a field element by frobenius coefficient. In many instances, these frobenius coefficients are 1, -1, however we resort to general multiplication in these cases.

The frobenius map power is almost always known at compile time, (and as far as I can tell only ranges between 0,1,2,3 in usage within the library), so we could switch this for a potential #206 optimization if we change the API to have different functions for different frobenius powers, or have a macro of some sort for computing this when the power is known at compile time.

This is potentially a notable speedup to the pairing

[Pratyush]

This is a good point, but I dunno how much we can save, because we only call the frobenius map a few times. I also don't know a good way to make the test for smallness cheap (at least for -1; for 0 and 1 it's easy).

[ValarDragon]

These are known at compile time, so theres no run time test needed with an API change. (The coefficient is a constant)

[ValarDragon]

Agreed that its probably not that notable on a second pass through, its called like 4 times for bls12