uneffective search for MOLS

[MathieuCossard]

When looking for mutually orthogonal Latin squares (MOLS), computation times are very different for sizes of the same order of magnitude, e.g.:

sage: time D = designs.mutually_orthogonal_latin_squares(2,303)
CPU times: user 962 ms, sys: 69 ms, total: 1.03 s
Wall time: 949 ms

sage: time D = designs.mutually_orthogonal_latin_squares(2,307)
CPU times: user 53.5 s, sys: 1.02 s, total: 54.5 s
Wall time: 54.5 s

The poor performance with 307 (or e.g. 409) is surprising, since the existence of a finite field F of order q enables to build q-1 MOLS: for a nonzero element in F, the a-th square is (aa_i+a_j)_{i,j}, where (a_i)_is an enumeration of elements in F (cf. http://www.les-mathematiques.net/phorum/read.php?34,1696222,1698540#msg-1698540).

Note: I ran into this with version 7.5.1. This might have been fixed since then.

Component: combinatorial designs

Keywords: MOLS, poor performance

Issue created by migration from https://trac.sagemath.org/ticket/26107

[pyrusbrawler64]

I am working on this, and hope to have a pull request ready soon.