Parsita

Magic classes (`TextParsers` and `GeneralParsers`) now restore the state of the global options (`whitespace`, `handle_literal`, `parse_method`) to what they were when the class started to run rather than to the defaults. This means that magic classes can be now be nested and magic classes will not clobber any manually set options.

Improvements to error messages and new utility functions:
* `TextParsers` error messages now provide line and character location rather than index into string
* Added `parsita.util` module with `contant`, `splat`, and `unsplat` helper functions
* Added a URL parser example to source

Addition of Python 3.6 support

Addition of a few helper functions:
- `eof`: a parser that matches the end of source
- `success`: a parser that always succeeds with a given value
- `failure`: a parser that always fails with a given message
- `Result.or_die`: returns the value on success and raises and exception on failure

Initial release of parser combinator library for Python:
- Fundamental parsers: literal, alternative, sequential
- Common parsers: optional, discard left and right, regex
- Repeated parsers: rep, rep1, repsep, rep1sep
- Forward declarations inside context
- Standard sequence context `GeneralParsers`
- Whitespace ignoring context `TextParsers`
- Examples: expression, JSON
- Complete unit test suite
- Basic tutorial