Towards sparse matrices with lift_std

[HechtiDerLachs]

This is a first take on #4925 . The issue on the Singular side (that lift_std should return a sparse matrix directly) has not yet been addressed. But these are the changes necessary to use sparse data types on the Oscar side.

@ederc : Maybe we want similar things for ideals? I think there lift_std still returns a dense matrix. But if we want this, I could work on those parallel changes, too.

For comparison:

julia> R, (x, y) = graded_polynomial_ring(QQ, [:x, :y]);

julia> F = free_module(R, 1000);

julia> I, _ = sub(F, [x*g for x in gens(R) for g in gens(F)]);

julia> @time coordinates(x*y*F[1], I)
  1.774690 seconds (16.61 M allocations: 386.731 MiB, 35.32% gc time)
Sparse row with positions [1001] and values MPolyDecRingElem{QQFieldElem, QQMPolyRingElem}[x]

[HechtiDerLachs]

Note: Doctests seem to fail for unrelated reasons.

[benlorenz]

Note: Doctests seem to fail for unrelated reasons.

Should be fixed by #5914.