Improve the conversion of Oscar/Flint polynomials to/from Singular

[hannes14]

Improve the conversion of Oscar/Flint polynomials to/from Singular: several problems

function (F::N_ZpField)(a::Nemo.fpFieldElem)
   characteristic(F) == characteristic(parent(a)) || error("characteristic does not match")
   return F(lift(a))
end

and related: - characteristic is Int in both cases, but compared as BigInt
- internal representation is the same: an Int between 0 and p-1
Singular has internal routines for conversion between flint (fmpq_poly_t,nmod_poly_t) and Singular (poly),
but unused here

[fingolfin]

Big question here: do we have examples of actual computations where the conversions really factor into the overall performance (and are not totally dominated by e.g. Gröbner basis computations, both in terms of time and space usage).

If so, then there is far more motivation to work on this, and to increase the priority...

So, if you have such examples, that matter to you, please provide them to us, and we'll really try to put this on the agenda.

---

That said, even without that, it is clear that our current conversion is not good, that can be checked with small examples.

To make progress on this, we need an expert on Singular internals (so basically @hannes14 :-) ) plus someone who knows Nemo / FLINT very well (so e.g. @thofma @lgoettgens @fieker or myself). Then such a tandem could sit down for a couple days to make this better (and my hope is that if we do this for a few types, the patterns become clearer and it won't be too much work. Ah well, we'll see)