Code optimization

[zStupan]

• Updated get_rules to accept algorithm as a string of the class name and set algorithm parameters from kwargs
• Replaced Dataset.feature_report() with a __repr__ method, which outputs detailed information about the dataset, including: Number of transactions, Number of features, Data types, min, max, categories of each feature.
• Optimized code for performance by:
  • Converting Feature and Rule classes to use __slots__ (this means constant attribute access and less memory usage)
  • Optimized the __post_init__ method to not use DataFrame.agg and using numpy arrays for storing the boolean mask of transactions containing the antecedent and/or consequent and then just summing those masks instead of subindexing transactions
• Fixed Yule's Q computation
• Updated CLI docs

[zStupan]

I'm still not sure if I got Yule's Q correct.. The Definition in that IEEE article gives a different result to this one

EDIT: I've decided to use the wikipedia definition, which is basically the same as the one in the IEEE paper, just more compact.