Implement tokenization via readtables

Note: This work is definitely not complete, functional or ready to be merged.

This is my first (published) pass at a readtable implementation. It isn't hooked into the pipeline yet but does show the general strategy and initial API. Per some previous discussions https://github.com/sweet-js/sweet.js/issues/535#issuecomment-242791944 I've intentionally implemented a very limited API (utility libs can build on these).

Readtables can currently be created as follows:

import { defaultReadtable } from 'src/readtable';
import CharStream from 'src/char-stream';

const newTable = defaultReadtable.extendReadtable({
  key: '@',
  action(stream) {
    return {
      type: 'Punctuator',
      value: stream.readString(),
      locationInfo: stream.locationInfo
    };
  }
});

const entry = newTable.getEntry('@');
const stream = new CharStream('@foo');
entry.action(stream);
// =>
// {
//   type: 'Punctuator',
//   value: '@',
//   locationInfo: {
//     filename: '',
//     line: 0,
//     column: 0,
//     position: 0
//   }
// }

The intention is to allow lookbehind of already read tokens. I'll have to implement this for '/' resolution. So the signature of action will be action(stream: CharStream, prefix: TokenTree?). TokenTree = Token | List<TokenTree>.

I have not implemented Racket style reader modes as I believe they can be build using the current primitives (i.e. by referencing other readtables in a closure).

As a note, I'm using some "private" fields (denoted by a _ prefix). I'm reserving the right to truly hide them in the future if there is a reason and efficiencies don't suffer too greatly. So their presence shouldn't be relied upon and shouldn't be part of the public API.

@disnet I'd appreciate some feedback as soon as you get the chance. I'm going to have a lot of time the next few days to hack on this and don't want to go too far down the wrong path.