Trial division in fmpz_poly_factor

[fredrik-johansson]

In our test suite (comprising qqbar etc.) something like 10% of all inputs to fmpz_poly_factor have a cyclotomic factor. We should investigate whether trial division by cyclotomic polynomials can be done quickly enough (e.g. mod 2^63 and with precomputed tables) to be worth it.

Other small linear and quadratic factors are also reasonably common, but perhaps not common enough for trial division.