Automunge

- found an opportunity for edge case troubleshooting printout
- associated with cases where a transformation function records a transfomation category in column_dict
- which doesn't have corresponding entry in process_dict
- now printout provides explanation
- note that in general Automunge includes prinouts for all identified potential error channels to support troubleshooting

- new mxab normalization option for max absolute scaling
- which quite simply divides data by max absolute value found within training set
- which returns a range not to exceed -1:1
- just trying to be comprehensive in normalziaiton options
- have seen this normalizaiton procedure referenced in a few places
- including Spark by Chambers Zaharia and ML Design Patterns by Lakshmanan et al
- just finally got around to it
- still recomend z-score normalization as default under automation for unknown distributions

- new datetime transform tmzn applied upstream of time stamp transform aggregations date/dat2/dat3/dat4/dat5/dat6
- defaults as a pass-through
- when timezone parameter passed to the aggeragate transform category, can designate desired time zone
- this doens't really impact sin/cos scalings
- primary benefit is for the business hours bin aggregator
- note pandas accepts kind of non-intuitive abbreviations for time-zones consistent with pytz.all_timezones

- improved convention for time transforms tmsn and time, which segregate time series data by time scale
- now if a particular time scale is not present in the training data that segment ommitted in returned data
- e.g. if a time series records calendar days but not clock times, hour/minute/seconds ommitted in retuned data
- we expect these type of scenarios are not uncommon in real world data sets
- also added inplace support for datetime transforms built on top of tmsc and time transformation functions
- also added inplace support for datetime transforms built on top of wkds and mnts transformation functions

- parameter support added to binning transforms with user specified buckets (bkt1,bkt2,bkt3,bkt4,bkb3,bkb4)
- now buckets parameter can be passed as percentages of range instead of set specific values
- eg with bucket boundaries in range 0-1
- in order to signal this option bcuket boundaries should be passed as a set instead of list
- eg for a set with range -5:15, buckets with exact boundaries could look something like [-5,0,5,10,15]
- and buckets as percentages could look like eg {0,0.25,0.50,0.75,1}
- which would give a consistent output
- partly inspired by a comment in book "Spark The Definitive Guide" by Bill Chambers and Matei Zaharia

- corrected processdict entry for DPnb
- which had incorrect True classification for inplace support
- resulting in intermediate column retention from upstream nmbr application
- now upstream nmbr is subject to replacement as intended