Use of fmpz ** exp for multivariate exponents

[fredrik-johansson]

The mpoly functions that deal with multiprecision exponents (e.g. fmpz_mpoly_get_term_exp_fmpz) use fmpz ** exp, requiring the creation of a vector of fmpzs as well as a secondary vector of pointers into the vector of fmpzs. What is the rationale for this and is it too late to change? We use fmpz * exp everywhere else in flint for vectors of integers, and it's inconsistent with the single-precision versions of the same function (e.g. fmpz_mpoly_get_term_exp_ui) which just take an vector of values.

[tthsqe12]

I don't like the fmpz** either. Some high level languages are not capable of forming an array of fmpz's as fmpz*. Probably this can be changed so there there is a _ffmpz version taking a flat array fmpz* and a _pfmpz version taking a fmpz** argument. This split already exist internally. Maybe @wbhart has something to say. As an interesting side note, there is no function for creating an mpoly from an array of coeffs and a matrix of exponents because we could not figure out what a matrix of fmpz's would look like from julia.

[fredrik-johansson]

I propose looking into deprecating the ** versions and renaming ffmpz -> fmpz. This is not urgent to fix, but it should help remove a bit of bloat.