Our current testing framework only really tests things "one way" in that we fixture up a figure, get plotly.js to process it, render the editor, and optionally check that the right event payloads fire on input, but we don't actually send those payloads back into plotly.js to close the loop. This prevents us from doing real end to end tests and allows bugs like the deletion one I hunted all of last week to evade automated tests.

Let's add some test utils to make end to end testing easier and then actually start building such tests!