BUG/REF: discrete fit_regularized uses print instead of warning

[josef-pkt]

print(message) in base.l1solvers_common

in writing a unit test for GeneralizedPoisson fit_regularized #3789, I get the warning "print" if penalization alpha is small. Given that it is a print I don't see a way to catch it in unit test noise.
I'm testing that low penalization has results close to unpenalized fit

>>> tt.res1.model.fit_regularized(alpha=alpha*0.1)
QC check did not pass for 5 out of 10 parameters
Try increasing solver accuracy or number of iterations, decreasing alpha, or switch solvers
Could not trim params automatically due to failed QC check.  Trimming using trim_mode == 'size' will still work.
...

[josef-pkt]

Do we need a special warning or do we use ConvergenceWarning?

postponing to 0.10,
AFAIR, qc check needs review and documentation, e.g. #2239