Automunge

- fixed improperly named variable

- changed dtype evaluation method for floatprecision
- replaced mnm0 category with mmor

- new automunge parameter floatprecision
- new root categories or10, om10, mnm0

- new infill option modeinfill, inserts most common value for a set
- modeinfill also works for one-hot encoded sets
- changed default option for exc2 category to mode

fixed small bug with standard deviation bins introduced in last update

- changed default automated time series processing to 'dat6'
- added special cases for all standard deviation and mean abs dev normalizations if ==0 then set = 1 to avoid div by 0
- changed all float to float32, all bool int to int8, all ordl left alone
(the thinking is that since the data is normalized we don't need as many significant figures, note that we're keeping this conversion inside the processing functions so we have option to later create alternate processing functions with different precisions)
- new processding function ord3, ordinal processing in order of frequency of occurance
- and ord4 (ord3 with minmax scaling)
- for logarithmic and powers transofrm, replaced missing data with 0 instead of mean (makes more sense)
- new featuremethod option 'report' to only return results of feature importance evaluation with no further processing
- fixed bug for assembly of ID set in postmunge