Fix/cost weight inconsistent

Creation of cost via abbreviation and constructor led to inconsistent cost weights, esp for those with default value != 1. (see first failing CI).

This MR fixes this problem and removes poorly documented default_weights != 1. as these should be handled more verbosely.

Possible side-effect: Some costs are rather rewards, therefore default cost will rise, e.g., with more goals fulfilled. But should then be just explicitly defined with weight < 0.