Automunge

- new feature engineering transformation category 'spl2'
- similar to splt rolled out in 2.64, but instead of creating new column identifier
it replaces categorical entries with the abbreviated string overlap
for example, the set of unique values in a column {'square', 'circle', 'spongebobsquare', 'turingsquared'}
would be transformed such that column only containd the set {'square', 'circle'}
- also defined a suite of sets of feature engineering functions that make use of this method
including 'txt3', 'spl2', 'spl3', 'spl4', 'ors2'
entry to assigncat in default automunge call for completeness

- added "versioning serial stamp" to automunge printouts
- new processing function 'splt':
compare unique values of categorical set strings, if portions of the strings have overlap,
create a new column indicator
for example, if a categoical set consisted of unique values ['west', 'north', 'northeast']
then a new column would be created idenitifying cells which included 'north' in their entries
(here for north and northeast)

- new postprocess_dict entries to support versioning:
'application_number', 'application_timestamp', 'version_combined'
- improved printouts for driftreport
- new driftreport metric:
'pct_NArw' tracks percent of cells subject to infill for NArw columns

- new postmunge option driftreport
- performs evaluation of data set drift for comparison between columns of the original training data passed to automunge(.) and corresponding columns of subsequent data passed to postmunge(.), prints report.

- simplified logic tests to evaluate default transformations categories

small updates:
- added support for edge case of user passing empty dictionaries to assigncat and assigninfill
- cleaned up a few errant comments