Improve performance in polynomial multiplication by using FFT

[batzor]

Currently, in univariate polynomial-by-polynomial multiplication, it uses the naive approach which takes $O(n^2)$. But if we utilize FFT, it can be reduced to $O(nlogn)$

tachyon/tachyon/math/polynomials/univariate/univariate_polynomial_ops.h

Lines 460 to 462 in ec92fa4

	static void DoMul(const UnivariatePolynomial<D>& a,
	const UnivariatePolynomial<D>& b,
	UnivariatePolynomial<D>& c) {

Reference: https://www.cs.toronto.edu/~denisp/csc373/docs/tutorial3-adv-writeup.pdf

[Dyslex7c]

Hi, would love to try this, can I? Is this still active?