formalize generative model; simulate fitness trajectories and check inferences

[matsen]

We are going to be developing machinery for detecting/quantifying changes in mutation fitness.

In such a setting, I think it's appropriate to set up simulations to benchmark our ability to do such work. I'm thinking something along the lines of

from this paper.

But to do that we need to more clearly formalize the generative model.