CI segfault on 32-bit Alpine / musl

[fredrik-johansson]

The change in randomization in commit f9764e5 revealed a segfault which occurs reliably on the Alpine Linux, musl, 32-bit (assert, x1.5) CI job. I'm unable to reproduce the error on a different 32-bit x86 setup (gcc 12, glibc 2.36 / debian 12).

As far as I can tell from printf debugging on the CI job, the segfault occurs at the end of iteration 586 of the test for gr_rsqrt_series when clearing a gr_poly of length 1 with nmod coefficients. That should obviously never fail, suggesting some earlier memory corruption.

[albinahlback]

I wonder if we should use musl on some 64-bit machine as well. If I recall correctly, they catch more errors.