This PR changes the renderProp API as described in #1863 (comment).

This is merely a suggestion at the moment and I'm happy to close it again if we settle on another approach or even keep the current one.

The motivation for "returning a function" is that the arguments to enzyme's renderProp won't get mixed with the arguments to the render prop function itself. This allows us to eventually add more options (like context) which we can apply to the newly created wrapper.

An example usage (formatted with Prettier) would look like this:

const wrapper = shallow(<App />)
  .find(Mouse)
  .renderProp("render")(10, 20)
  .find(Mouse)
  .renderProp("foo")(3, {
    long: {
      yeah: true
    }
  })
  .find(Bar)
  .renderProp("x")();

Sidenote

At this point, one might wonder what the advantage of wrapper.find(Foo).renderProp('render')() over wrapper.find(Foo).prop('render')() is, and that's a fair question.

When using renderProp, the render props' return value gets nested into a new wrapper which enables this convenient chaining shown above - whereas wrapper.find(Foo).prop('render')() simply returns the React element without an enzyme wrapper. More about this is hinted at in this article.