Generalize the library rubric implementation

Currently, the library supports 23 rubrics in all. However, the library should be made more generic to support additional rubrics that a user might think off. This would be that they would supply additional input arguments like rubric-name, rubric-description, likert-scale rating definitions, etc. This should be supported in both pointwise and pairwise LLM-as-a-judge evaluation settings.