Create better alternatives to Csv.ProcessFoo methods

[airbreather]

Csv.ProcessFoo methods don't scale very well. Every new type of input requires all the logic for scanning through the input to live in this file, and it doesn't compose very well at all.

Furthermore, all the parameters for how to process that input (e.g., buffer sizes, delimiters, ignore UTF-8 BOM or not, pooling, etc.) need to be specified as method parameters, in a particular order, on the same method call. This makes it hard to add parameters in the future.

This can all be solved by moving to an object model that serves the same purpose as the Csv class has served to-date, but with more options and an interface that's easier to extend.

[airbreather]

This was done in a89f99f.