Automunge

- new processing functions nmrc/nmr2/nmr3/nmcm/nmc2/nmc3/ors7/spl7
- nmrc: parses strings to return any numeric entries, entrisd without number suject to infill
- nmrc: if multiple numbers found in entry returns the number with most characters
- nmrc: does not recognize numbers with commas, for instance '1,200.20 ft' returns as 200.20
- nmr2: nmrc with z-score normalization applied to returned numbers
- nmr3: nmrc with min-max scaling applied to returned numbers
- nmcm: similar to nmrc, but recognizes commas, for instance '1,200.20 ft' returns as 1200.20
- nmc2: nmcm with z-score normalization applied to returned numbers
- nmc3: nmcm with min-max scaling applied to returned numbers
- ors7: similar to ors6 family of derivatoins as demonstrated in last essay and incorporates nmr2
- spl7: same as spl5 but recognizes string character overlaps do0wn to single character instead of minimum 5
- new NArowtype 'parsenumeric' identifies rows without parsed number entries as subjhec tot infill
- new NArowtype 'parsenumeric_commas' identifies rows without parsed number entries with commas as subject to infill
- new NArowtype 'datetime' identifies rows without datetime entries as subject to infill
- changed name of NArows(.) function to getNArows(.)
- new support functions for these methods is_number(.), is_number_commas(.), parsenumeric(.), parsedate(.)

- replaced the default trasnforms for powers of ten encoding with 'pwr2' (one-hot) and 'por2' (ordinal)
- powers of ten now encodes both positive and negative values as 'column_10^' and 'column_-10^'
- powers of ten one-hot encoding now defaults infill as no activations
- powers of ten ordinal encoding now matches test set potential set of values to train set
- powers of ten ordinal encoding now defgaults infill as distinct encoding (0)

- simplified bulk exclusion from processing with new powertransform option
- if user passes powertransform = 'excl', columns not explicitly assigned to a root category in assigncat will be left untouched
- or if user passes powertransform = 'exc2', columns not explicitly assigned to a root category in assigncat will be forced to numeric and subject to default modeinfill
- made a special case for 'excl' category in general in that columns processed with this (pass-through) function are now returned without a suffix appender to the column header
- found and fixed missing application of floatprecision to labels set if included in test data

- found and fixed error in dxd2 transform (number of rows offset was inconsistent with documentation)
- new transform 'sqrt' applies square root transform
- new NArowtype 'nonnegativenumeric', similar to 'positivenumeric' but allows 0 values
- new NArw root category NAr4 based on nonnegativenumeric

- updated TrainLabelFreqLevel (class imbalance levelizer) function for support of numerical data levelized basis on powers of 10 for floats <1 (eg 10^-1, 10^-2, etc)

- updated splt categories to base search of max string length in unique values
- updated methods for splt, spl2, spl5 to account for bug when overlaps exceed hard coded search length. Now method bases string overlap search on max length of unique values.
- updated length of min evaluated overlap length to match READ ME