Dynamic determination of convergence – introduce `has_converged`

[kellertuer]

Currently there exists indicates_convergence as a static way to a stopping criterion whether it has the possibility to report convergence (i.e. a maxiter does not, a tolerance on the gradient does).

This could be enhanced by a second (dynamic) variant, m<ybe called has_converged that would be more parallel to get_reason, that is it would

• for single criteria report true if they indicate_convergence and is “active” that (have a reason that is not empty)
• for Any (|) it would check whether one of its “active” criteria is a criterion that indicates convergence
• for All (&) it would check whether all of its criteria are “active” and at least one indicates convergence.

This came up as a feature idea in JuliaMolSim/DFTK.jl#1105