Quantile Binning Refactor

[Schobs]

Discussion for TODO items on the QBin Refactor.
Currently @ PR here .

TODO:

• boxplot display settings as params in interpret/uncertainty_quantiles functions.
• Circular import issue in evaluate_correlations(). Line 98 of kale/evaluate/similarity_metrics.py
• Refactor of evaluate/uncertainty_metrics for repeated code blocks.
• Refactor of interpet/uncertainty_quantiles for repeated code. Possibly make Boxplot class.
• Consider replacing custom Jaccard with sklearn learn jaccard.
• There are repeated for loops and fragments, could you implement them as functions/classes to reuse?
• warnings.filterwarnings("error") at examples/landmark_uncertainty/main.py
  • Do you have to put this between imports?
  • Also, is ignoring errors a good practice?
• Improve Code Coverage for:
  • similarity_metrics.py - saving figs in evaluate_correlations() crashes on Github Actions
  • uncertainty_metrics.py - saving figs in fit_line_with_ci crashes on Github Actions