Path parser based on string:tokens/2 rather than neotoma peg parser.

[tsharju]

This branch replaces neotoma parser with a custom parser based on string:tokens/2. All tests pass. The only difference is that the characters used in property names is not limited in any way.

A quick timing test shows quite significant performance boost.

Master branch:

timer:tc(props_path_parser, parse, ["test.test[2].test"]).
{139,
 [{prop,<<"test">>},
  {prop,<<"test">>},
  {index,2},
  {prop,<<"test">>}]}

This branch:

timer:tc(props_path_parser, parse, ["test.test[2].test"]).
{9,
 [{prop,<<"test">>},
  {prop,<<"test">>},
  {index,2},
  {prop,<<"test">>}]}