Improve comparing Singular and Julia/Nemo coefficients

[fingolfin]

Consider this:

julia> a_nemo, b_nemo, c_nemo = Nemo.ZZ(5), Nemo.ZZ(7), Nemo.ZZ(3)^100;

julia> a_sing, b_sing, c_sing = Singular.ZZ(5), Singular.ZZ(7), Singular.ZZ(3)^100;

julia> @b $a_nemo == $b_nemo
4.013 ns

julia> @b $a_nemo == 5
3.705 ns

julia> @b $a_nemo == $b_sing
138.743 ns (5 allocs: 96 bytes)

julia> @b $a_sing == $b_sing
52.994 ns (3 allocs: 48 bytes)

julia> @b $a_sing == 5
311.557 ns (11 allocs: 176 bytes)

Especially the last one is really bad, comparing a Singular integer to a Julia machine int should not require so many allocations.

---

Of course the above is just with n_Z, but it also affects n_Q and finite field elements and probably others, to varying degrees.

[fingolfin]

Of course similar things apply to <

[fingolfin]

Oh, forgot to CC @hannes14 @fieker