Automunge

- new processing root cateogry family trees: or11 / or12 / or13 / or14
- or11 / or13 intended for categorical sets that may include multiple tiers of overlaps and include base binary encoding via 1010 suppplemented by tiers of string parsing for overlaps using spl2 and spl5, or11 has two tiers of overlap string parsing, or13 has three, each parsing returned with an ordinal encoding sorted by frequency (ord3)
- or12 / or14 are comparable to or11 / or13 but include an additional supplemental transform of string parsing for numerical entries with nmrc followed by a z-score normalization of returned numbers via nmbr

- replaced postmunge returned object finalcolumns_test with postreports_dict
- postreports_dict contains results of optional feature importance evaluation, drift report, and list of final columns, may be used for future extensions as well
- see postreports_dict['featureimportance'] / postreports_dict['driftreport'] / postreports_dict['finalcolumns_test']
- updated printouts for automunge and postmunge associated with feature importance and postmunge drift report, postmunge printouts now tied to printstatus parameter
- fixed bug in automunge and postmunge for processing of dataframes with non-range integer index

- new processing function 'UPCS' converts categorical string sets to uppercase strings
- such as for consistent encodings if same entry included with upper and lowercase characters
- (e.g. 'USA' and 'usa' would be consistently encoded)
- new suite of categorical processing functions incorporating UPCS upstream of encodings:
- Utxt / Utx2 / Utx3 / Uord / Uor2 / Uor3 / Uor6 / U101
- comparable to text / txt2 / txt3 / ordl / ord2 / ord3 / ors6 / 1010

- new processing functions for mean normalization => (x - mean) / (max - min)
- mean: basic mean normalization
- mea2: mean normalization coupled with a z-score normalization
- mea3: mean normalization coupled with standard deviation bins

- new processing functions nmr7/nmr8/nmr9/nmc7/nmc8/nmc9
- comparable to functions nmrc/nmr2/nmr3/nmcm/nmc2/nmc3
- but make use of new method whereby test set entries are string parsed only in cases where those unique entries weren't found in the train set

- new processing functions nmr4/nmr5/nmr6/nmc4/nmc5/nmc6/spl8/spl9/sp10 (spelled sp"ten")
- comparable to functions nmrc/nmr2/nmr3/nmcm/nmc2/nmc3/splt/spl2/spl5
- but make use of new assumption that set of unique values in test set is same or a subset of those values from the train set, which allows for a more efficient application (no more string parsing of test sets)