Fix issue310

[jerebertho]

This PR fixes issue #310

1. When the DRL Gröbner basis has a linear form x_n - a, there was issue in the change of order step,especially to compute the minimal polynomial of x_n. This could yield problems when computing the parametrizations.
2. The degree of the minimal polynomial of x_n and that of its squarefree part were tracked several times in the change of order algorithm. This has been simplified.
3. The computation of the inverse of the Hankel matrix, for computing the parametrizations, was wrong in a rare case that happens with probability 1/p, where p is the characteristic of the field. Since in issue Segmentation fault with a simple example #310 , the characteristic was 7, the problem was quite frequent.