Automunge

- assigninfill now supports passing column headers of dervied columns in addition to source columns
- thus enabling user to assign distinct infill methods for different columns dervied from the same source column
- (assigning source column headers are applied to all downstream columns in aggregate, and then passing a derived column header will override)
- replaced a function call associated with infill application in the postmunge(.) function for improved efficiency

- ML infill now trains models even if automunge(.) passed sets have no infill points
- for potential use in postmunge(.)
- (had meant to roll this out a while ago)

- improved memory overhead of ML infill by consolidating seperately named sets in a few data conversions between pandas and numpy
- improved memory overhead of feature importance evaluation in similar fashion
- improved memory overhead of PCA dimensionality reduction in similar fashion

- found and fixed small edge case for automunge(.) powertransform evaluation function under automation
- found and fixed small typo for category identifiers logged for bnwK, bnwM, bnKo, bnMo
- found and fixed small bug for driftreport assembly in postmunge(.)

- found and fixed silly bug for ML infill
- (had fixed this while back and just realized that update got set on hold before rolling out)
- found and fixed bug for postmunge(.) feature importance corresponding to application of Binary parameter activation in automunge(.)

- new transform category 'ptfm', similar to 'eval'
- the intent is that the (modular) data property evaluation function for automation may have a few distinct default bases activated by the powertransform parameter:
- a regular default evaluation when powertransform = False, and a seperate (potentially more elaborate) default evaluation basis when powertransform = True
- or the evaluation may be turned off such as by passing powertransform = 'excl' (which means unassigned columns are left unaltered)
- the purpose of the 'eval' and 'ptfm' options are for cases where the automated evaluation is set to one of these bases (regular, powertransform, off), a user can still assign to distinct columns one of the other bases for evaluation
- added support for 'eval' and 'ptfm' options to label columns
- a little bit of code cleanup in automunge(.) function