WIP: add ability to work with mixed trees

[systemallica]

• Added morfist as a mixed tree learning algorithm, capable of predicting nominal and numerical variables using a single model. Based on RandomForests.
  • New parameter added to Mercs: ‘mixed_algorithm'.
  • New configuration object: ‘mixed_config'.
  • New global config object: ‘mixed'.
  • Updated selection algorithm so that it selects mixed input/outputs if morfist is selected as learning algorithm.
  • Updated the induction algorithm to add mixed models.
  • Updated CanonicalModel.py to handle mixed models.
  • Update the inference algorithm to correctly retrieve nominal values in mixed trees.
• General code improvements and cleanup.
• Removed unused code and files.
• Improved documentation.
• Added some unit tests for mixed trees.
• Added 'decision-tree-morfist' as project dependency.
• Added poetry as dependency manager.
• New CI strategy(using GitHub Actions).
• Simplified deployment using Poetry.