Updates convergence rules to keep current functionality but allow them also work with negative values (useful in principle if we want to use a feature which reaches a minimum at the desired optimum -- we can just maximize the negative) and cope if parameter values increase during optimization (which is unusual but can happen). Also splits the behaviour currently activated by favour_narrower_parameter in two: continuing until a significant decrease is enabled now by try_narrower_values in the calibration field, while choosing the narrowest parameter within a certain deviation from the maximum is enabled by favour_narrower_optimum in the calibration field.
Updates convergence rules to keep current functionality but allow them also work with negative values (useful in principle if we want to use a feature which reaches a minimum at the desired optimum -- we can just maximize the negative) and cope if parameter values increase during optimization (which is unusual but can happen). Also splits the behaviour currently activated by favour_narrower_parameter in two: continuing until a significant decrease is enabled now by try_narrower_values in the calibration field, while choosing the narrowest parameter within a certain deviation from the maximum is enabled by favour_narrower_optimum in the calibration field.