Ruleminer

- Improved performance of evaluation (antecedent and consequent are calculated only once)
- Added corr-function to calculate correlations
- Refactored code
- Added some preparations for logging rule evaluations

- Improved logging info of evaluation of rules

- Fix in evaluation of unequal operators in combination with tolerances
- Fix in evaluation of equal and unequal operators when writing debug info

- Added exact-function to grammar and parser to ignore tolerance of specific part of expression
- When exponentials like x ** y are used, for x max(0, x) is substituted to prevent imaginary numbers (as x can be negative)
- The tolerance dict can now contain None values:
- if params={'tolerance': {"A": None', 'default': { ... }}} then tolerance is not applied to column A and other columns are set to 'default'
- if params={'tolerance': {"default": None, "A": { ... }}} then tolerance is not applied as default except for column A
- the keys of the tolerance dict can contain regular expressions. They are matched with columns when converting the expressions
- Added regex to dependency in pyproject.toml

- Added mean and std functions to grammar and parser
- Changed parameter name 'evaluate_quantile' to 'evaluate_statistics' to cover mean and std
- Changed subst parameter definitions: substr(x, a, b) is now translated to x.str.slice(a-1, a+b-1)

- Added multiply and divide functions when left and right side both contain columns (see usage documentation)
- Changed grammar to separate plus, minus on one hand and multiplication, division on other hand into separate groups (needed for first point)
- Added debugging information on lower and upper bounds when evaluating (in)equalities