Tests on number of iterations are sensitive to randn

[rfourquet]

A PkgEval run for a PR which changes the generated numbers for randn! indicates that the tests of this package might fail in Julia 1.5 (and on Julia current master). Apologies if this is a false positive.
cf. https://github.com/JuliaCI/NanosoldierReports/blob/7de24e455342298cbef56826b5827f0d7640d2c1/pkgeval/by_hash/b89e35c_vs_098ef24/logs/ProximalAlgorithms/1.5.0-DEV-71a4a114c2.log

[nantonel]

@lostella I remember we decided to relax the tests on the number of iterations (I couldn't find the issue though).

Looks like they were not relaxed enough? The order look very similar, so it is definitively that:

[...]
  Expression: it_dys <= 1900
   Evaluated: 2363 <= 1900
[...]
  Expression: it_dys <= 2600
   Evaluated: 2877 <= 2600
[...]
  Expression: it_afba <= maxit
   Evaluated: 530 <= 500

I wonder if these tests could be replaced by something more robust...

[lostella]

@nantonel I think at this point it would be great to replace these tests with some fixed problem, instead of randomly generated...

[nantonel]

I see that most calls to randn don't use a fixed random generator likerandn(rng, ComplexF32, (2, 3)).

That should avoid pasting numbers like in the LBFGS tests unless random generators are changed which I don't think happen very often.

[lostella]

@nantonel #38 fixes this