Exact Rational Values about arb

[fredrik-johansson]

This only works for integers because integers are a subset of floating-point numbers. Doing it with rationals would require a union type. I'm not sure if this would be worth the effort to implement at the moment, considering that there are workarounds.

Where it's beneficial to preserve exact rationals, I've generally just made two versions of a function: one that takes an fmpq_t argument and one that takes an arb_t.

In some other places, I've allowed the input p / q to be passed as two separate arb_t variables, doing arithmetic with numerators and denominators separated.