Itertable

- Fix minor issue with exception strings.

API Improvements
- Python 3 support (2)
- Use [requests](http://python-requests.org) for `NetLoader` (#3)
- Read and write support in all *FileIO classes (`CsvFileIO`, `JsonFileIO`, `XmlFileIO`, and `ExcelFileIO`)
- Comprehensive test suite
- More robust `pickle` support
- Complete docs on [extending wq.io](http://wq.io/docs/custom-io)

API Improvements
- Add `Exception` subclasses to report common error cases (e.g. `LoadFailed`, `ParseFailed`, `NoData`) - see heigeo/climata-viewer5
- Make `TupleMapper` instances pickleable - see heigeo/climata-viewer8

API Improvements
- Formalize notion of a `nested` IO, or an IO where each record contains a pointer to another IO. This concept is used extensively by the [climata](https://github.com/heigeo/climata) library, for example in [climata.acis.StationDataIO](https://github.com/heigeo/climata/blob/master/climata/acis/__init__.py#L120).
- Added a utility function, `flattened()`, that converts nested IOs into single-level IOs through denormalization
- Other minor improvements

API Improvements
- Don't crash when attempting to parse a non-existent (or empty) CSV file (5)
- Restore CSV `field_names` functionality for files without headers (6)

API Improvements
- Cache index in `BaseIO` to speed up lookups when used as a `dict`
- Use an array instead of a generator in `CsvParser`