Ruleminer

- Fix in grammar when parsing mathematical expressions in function parameters
- Fix for applying tolerances

- Added: 'not in' as comparison operator
- Added: parameter 'apply_rules_on_indices' (default True) that allows indices to be included as columns when generating and evaluating rules (if set to False then the data DataFrame is not changed)
- Output: dtypes of results DateFrame are enforced
- Performance: changed the internal equal and unequal functions to improve performance for columns with string, bool and datetime64_ns
- Logging: change to logging when all are not applicable

- Improved parsing performance

- Changed tolerance function to be able to evaluate string values
- Small fix when collecting results

- Refactoring
- Created separated CodeEvaluator class
- Created separated RuleParser class
- Streamlined parentheses process (resulting expressions have same parentheses as original)
- Performance
- Deleted concats when collecting rules and results
- Parameter 'results_datatype' can now be a dict that contains per column a list of results

- Parser of list comprehension now accepts a wider range of expressions
- Fix when applying tolerances when denominator in rule expression is negative
- Shortened internal function names