add parameter to "Dummify" feature to use top n values by rank or frequency

_From @rayidghani on December 22, 2016 17:50_

When creating dummy variables (or counts), allow option to say "Create dummy variables or counts for the top n most frequent values from a given field" instead of requiring user to pass values for the top n individually. We can have the default behavior of creating an "other" field/feature

_Copied from original issue: dssg/collate#34_

The issue names the Dummify feature but no file, test, or entry point. Locate the Dummify implementation and its current handling of supplied values, then trace how dummy variables or counts are generated. Done means supporting top-n selection by rank or frequency and the stated default "other" field or feature.

Written by the indexing model from the issue text.