Enable rest props to pass to editor

[webOS101]

This allows for passing props into the editor from the wrapping app.

[kitten]

I'm not sure, I want to allow passing arbitrary props through the element through for now. And if so, it'll need to correctly append classnames, and proxy events in the callbacks :/

[webOS101]

I can add proxy event handling and classname handling, if desired.

[webOS101]

If you prefer a more surgical approach, for my use case I need blur/focus event handlers passed through.

[webOS101]

className was already being merged correctly in the editor. I added code to forward the events that were being handled internally by the editor. I did not add a handler for the onChange event in LiveEditor, but could add one if desired. Lastly, I made sure to delete any children passed in props, it didn't seem like that would be a good idea to pass through. Interestingly(?), contentEditable is a prop on the editor but there was no way to control it before.

[kitten]

Ok, last request. Otherwise lgtm :)

[kitten]

This is a rather non obvious v8 perf hog :) can you just manually set code to undefined for presentations below? Also you don't need to do anything about children

[webOS101]

Interesting. I don't think the lifetime of rest is such that it will be affected by the V8 issue, but I shall make the change requested and leave children alone.

[webOS101]

Unfortunately, setting it to undefined does not have the desired effect. It's still a property on rest and causes a React warning: Unknown prop code on <pre> tag. I removed the delete on children and the unneeded semicolon. The only other option is destructuring out code and ignoring it, but I don't like unused parameter warnings much.

[webOS101]

If you're interested, here's the React discussion on undefined vs delete and that warning: facebook/react#7266

[kitten]

@webOS101 Interesting. Seems I forgot about this behaviour due to the future shift in React towards not filtering anymore :) Let's get this merged then

[webOS101]

FWIW, I did a benchmark test of rendering 500 components using destructuring and again using delete. The difference, on average, was less than 1ms per run, with the destructuring being somewhat faster.

[webOS101]

whoops, I labeled my tests backwards. delete was faster.