Have gr_divexact over a gr_poly ring dispatch to gr_poly_divexact

[fredrik-johansson]

This should somewhat speed up fraction field arithmetic over gr_poly rings, among other things.

Sample benchmark:

#include "gr.h"
#include "gr_poly.h"
#include "profiler.h"

int main()
{
    gr_ctx_t R, Rx, Rxy, Rxyz;
    gr_ptr f, g, h;

    gr_ctx_init_fmpzi(R);
    gr_ctx_init_gr_poly(Rx, R);
    GR_MUST_SUCCEED(gr_ctx_set_gen_name(Rx, "x"));
    gr_ctx_init_gr_poly(Rxy, Rx);
    GR_MUST_SUCCEED(gr_ctx_set_gen_name(Rxy, "y"));
    gr_ctx_init_gr_poly(Rxyz, Rxy);
    GR_MUST_SUCCEED(gr_ctx_set_gen_name(Rxyz, "z"));

    f = gr_heap_init(Rxyz);
    g = gr_heap_init(Rxyz);
    h = gr_heap_init(Rxyz);

    GR_MUST_SUCCEED(gr_set_str(f, "(1+I+x+2*y+3*z)^10", Rxyz));
    GR_MUST_SUCCEED(gr_set_str(g, "(1+I-x+2*y-3*z)^10", Rxyz));
    GR_MUST_SUCCEED(gr_mul(h, f, g, Rxyz));

    TIMEIT_START
    GR_MUST_SUCCEED(gr_divexact(f, h, g, Rxyz));
    TIMEIT_STOP
    TIMEIT_START
    GR_MUST_SUCCEED(gr_gcd(f, h, g, Rxyz));
    TIMEIT_STOP
}

Before:

cpu/wall(s): 0.00129 0.00129
cpu/wall(s): 0.00864 0.00863

After:

cpu/wall(s): 0.000745 0.000744
cpu/wall(s): 0.00825 0.00825