Automunge

- new parameter accepted for numeric scaling options nmbr/mnmx/mean/retn/DPrt as stdev_cap
- used to specify a cap/floor based on a specified number of standard deviations from the mean
- defaults to False, only inspected when prior parameters cap/floor are not specified
- hat tip for the inspiration to ACM paper "Automating Data Science" by Tijl De Bie, Luc De Raedt, José Hernández-Orallo, Holger H. Hoos, Padhraic Smyth, and Christopher K.I. Williams

- found an edge case not covered by rollout validations associated with inconsistent data between train and test sets for 1010 transform
- new convention that majority str(number) sets under automation default to numeric encoding
- this edge case helped me recognize that needed a convention, there is an arguement to be made for treating this case as categoric as well, settled on this approach just to have one
- also resolved the edge case by new data structure in case user prefers the alternate approach
- also found and fixed edge case for noise scaling in DPmm/DPrt associated with zero division

- new infill option supported as interpinfill, available for assignment in assigninfill or through processdict defaultinfill
- interpinfill is build on top of pandas interpolate method, and simply applies linear interpolation between entries
- as may be preferred over adjinfill for time series data

- qttf subsample parameter was causing halt channel due to improper float type for default value, now case as int to resolve

- added a workaround for halt scenario in autoinference for cases where inference was attempted without a trained model

- running some validations on automodel and autoinference and found a few alignments needed to support running autoML types different than those used for ML infill
- new final model training pipeline demonstration based on XGBoost tuned with Optuna in read me as Final Model Training
- upgraded the final model functions automodel and autoinference from "experimental" to "Beta"