Adds two core blocks beyond generic blocks

Adds image and quote blocks.
Meant for experimenting with possible parse ideas.

This directly embeds certain core block types into the spec parser. The
benefit is that we have tight coupling and integration of those block
types. The cost is that we have tight coupling and integration of those
block types.

There's a bit more flexibility in integrating this higher-level parser
into the string parser, but I think it may be just as easy to include a
JS pass over the parse tree to convert generic into more specific nodes.

This is pretty verbose now but I could imagine a basic DSL (or an
existing one) to compare tree structure and (choke) load data from HTML

```
[
  'core/image',
  or( !! caption,
    [ [ 'img', { src: [ 'url', types.url ] } ] ],
    [ [ 'figure', [
        [ 'img', { src: [ 'url', types.url ] } ],
        [ 'figcaption', 'caption' ]
      ]
    ] ]
  )
] >>> {
  type: 'core/image',
  url: '...',
  caption: [
    // subtree inside of caption tag, maybe blocks
  ]
}
```

^^^ that's just playing around, but the idea could be used to both load
and generate the serialized data