Support non-UTF-8 encodings and fail gracefully with unexpected encoding

[chrispomeroyhale]

A customer was trying to import a Windows-1252 encoded document and the parser gets stuck in a loop when finding a bad encoding. For privacy reasons I will not post the particular document.

Two parts:

1. At the very least, we should gracefully handle documents with unexpected encodings so that client apps don't hang and can display an error.
2. Ideally support multiple encodings. Although we may expect modern CSV to be UTF-8 encoded, the whole idea of the CSV Dialect is interoperability with existing documents in their various flavors.