Add "data" methods to Model

[pwalsh]

Description

The model should be immediately usable for common data conversion. Add public methods that transform a stream/iterator, and a row of data.

For simplicity, the API expects a iterable of arrays (or an iterable which is a single row) where values are either:

• a unicode string (string on Py3, unicode on Py2)
• the type being cast to (already transformed)

A later iteration could accept a file path or a raw CSV stream, but to do this we should first solve the issue of a common tabular data reader that can be used in MessyTables or GoodTables (frictionlessdata/tabulator-py#1). Then, we can plug that in here.

Tasks

• Method for converting a single row
• Method for converting an iterable and yielding rows
• Methods could support a strategy argument which is one of cast or test, which controls the response values as per the cast and test methods of types
• The method that handles iterables should support a fail_fast boolean. If true, raise on first error, if false, collect errors and return an error iterator when iteration ends (@pudo mentions colander and its grouping of errors here... worth checking)
• Tests