- `user_repr` is now a method
- Fix for python 2.7
- excepts -> expected

Note: This refactor opens the door for implementing a ContextualLexer for Earley.
But unlike the existing one for LALR, it will have to be computed at runtime,
rather than ahead of time.

This is an additional response to #738.

I also often came across this when unexpectedly having a newline, which would previously result in the message being split up across two lines. This PR makes it easier to identify exactly which special character is causing problems.

make _all_terminals a dict
added raw attribute to Pattern
rename nice_print -> user_repr

rename LexerConf.terminals to LexerConf.tokens
Make Exception message generation lazy
Made a few classes new-style

fixes #663

Add:
- missing imports
- __version__ variable

Additionally regenerated json parser example

The recently added VisitError is the only lark exception not inheriting from LarkError,
which makes it painful to catch for client application.

* All exceptions are now under exceptions.py
* UnexpectedInput is now superclass of UnexpectedToken and UnexpectedCharacters,
  all of which support the get_context() and match_examples() methods.