support conditional rules #2
Description
It has become clear that many real-world use cases involve heterogenous data rows -- in short, data which require that the rules be applied selectively rather than to every row in the CSV.
To do this requires filtering rows, itself a fairly trivial problem (an initial attempt is available in the 'row-filters' branch), but this also raises some deeper questions about the optimal structure of the schema file. It may be beneficial to flip the schema on its side, making the column the first-level key in the rules dictionary, rather than continuing with the earlier decision (taken primarily for the sake of brevity) to have the rule be the first level key.
See also http://martinfowler.com/articles/refactoring-adaptive-model.html for a discussion of coding logic into configuration files.