Ruleminer

- 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

- Added dtype=float to np.sum operations to prevent divide by zero exceptions
- In the results a row with not applicable is only included is there are no exceptions and no confirmations
- Added days / months / years functions that return the number of days / months / years of an expression (the difference between two timedate64 columns)

- Fix for SUMIF when data DataFrame contains pd.NA's
- Added date functions to extract day, month, quarter and year from a column
- Added date functions day_name, month_name, days_in_month, daysinmonth, is_leap_year, is_year_end, dayofweek, weekofyear, weekday, week, is_month_end, is_month_start, is_year_start, is_quarter_end and is_quarter_start

- Added 'not applicable' as a default metric to be calculated (n - confirmations - exceptions)
- Restructured the parsing of parentheses (deleted some unnecessary pyparsing.groups)
- Fix for unnecessary parentheses around power function (**)
- Fix for necessary parentheses around mathematical expressions
- Adapted all unittests accordingly