0.0.2: smaller parser but parses more

This cycle was rather big, a lot of changes has been made
to both reduces the parser size and improve parsing correctness.

Let's start off with the parse statistics against the Nim compiler
source code + nimble:

Total parses: 3220;
successful parses: 3147;
failed parses: 73;
success percentage: 97.73%

The parser size is now 118MiB, which unfortunately is still too big
for GitHub, but we are getting there.

Below are the changes in detail:

Breaking changes

• Symbolic operators are exposed as syntactic nodes (#12)
• Call arguments are now stowed in an argument_list node (#15)
• Escape sequences for strings and char now have their own nodes (#13)

Parsing improvements

• Dot generic calls (eg. x.foo[:T]) are now fully supported
• Function with an empty body and only comments can now be parsed
• Complex expressions are now supported for enum values
• when in object declarations can now have fields declared on the same line
• Tuple-with-var return values are now correctly parsed
• Cast expression now supports colon expressions
• Block comment contents no longer erroneously set the parsing indent

Regressions

• Unicode operators support has been disabled to prevent parser size
  from growing uncontrollably
• Some parsing has been relaxed and "invalid" structures might be parsed
  successfully. See corpus/deviations.txt for known deviations from
  the Nim compiler.