kkr_scf workflow does not recognize stagnating convergence

[PhilippRue]

Especially for GGA calculations convergence might get stuck above the chosen convergence criterion. Then the kkr_scf workflow fails to recognize that a calculation has reached 'soft'-convergence already and resets to the last 'properly successful' calculation before submitting the next calculations (possibly indefinitely). This has, for example happenend here:

It would be better if KKR recognized such soft convergence and this could then be passed on to the kkr_scf workflow which should then stop (possibly with 'soft' convergence marked).