lark.lark didn't parse grammars that didn't have a trailing newline in the file, but lark itself has no problem parsing those grammars
this change makes lark.lark more accurate in reflecting what grammars are parsed by lark.

using `SIGNED_INT` instead of `INT` more closely matches the the grammar used by lark in `load_grammar.py`

The result is a much smaller file.

Add:
- missing imports
- __version__ variable

Additionally regenerated json parser example

- Refactored lexer interface into LexerConf
- Lexer now compiles regexps only when used (especially useful for ContextualLexer)
- Lexer now doesn't validate on deserialize (noticable speedup)

TODO: Add docs