clarify the documentation of `gr_poly_resultant`

[mezzarobba]

According to the current text, the “sylvester version [...] is useful over inexact rings and as a fallback for rings without division”, but “no algorithm has been implemented that is appropriate for integral domains”. This seems self-contradictory to me, though it is probably just me not understanding what is meant by “appropriate” here.

[fredrik-johansson]

Appropriate for integral domains would be to use the subresultant Euclidean algorithm since this has better complexity than a generic determinant computation. Likewise for gr_poly_gcd.