Improve expectation() operator by using Expectation Programming?

[J Heitzig]

Having read the paper https://arxiv.org/abs/2106.04953, I wonder if it would be difficult to implement their approach to improved expectation estimation to WebPPL. Any thoughts? If you think it appears not too difficult, I could give it a try...