Building morphisms of graded modules takes unexpectedly long

[HechtiDerLachs]

Try the following code:

S, (x, y, z) = graded_polynomial_ring(QQ, [:x, :y, :z])
m = ideal(S, gens(S))
S1 = graded_free_module(S, [0])
g = gens(m^29)
f = [f*S1[1] for f in g]

@time I, inc = sub(S1, f);

@time Oscar.SubModuleOfFreeModule(S1, f);

On my machine I get

julia> @time I, inc = sub(S1, f);
  0.160066 seconds (2.22 M allocations: 98.650 MiB, 52.27% gc time)

julia> @time Oscar.SubModuleOfFreeModule(S1, f);
  0.000013 seconds (10 allocations: 1.141 KiB)

which is a significant differences. In fact, using sub in bigger examples killed me.

Using ProfileView revealed that most of the time is spent in inferring the degree of the generators for the domain of the inclusion map inc. This looks to me as if there's room for improvement.

@jankoboehm ?