Make counterfactual transportability more like surrogate outcomes paper

[djinnome]

• Create a counterfactual-transportability subdirectory under notebooks
• Split out the 3 use cases into separate notebooks
  • Unconditional counterfactual transportability (COVID)
  • Conditional counterfactual transportability (HIV)
  • Using multiple source populations to identify query in target population (Climate change)
• Add a README that summarizes the notebooks and how they differ

[djinnome]

Narrative

1. Start with a counterfactual query that cannot be identified from observational data in the target domain
2. Introduce transportability:
   • If we can identify the counterfactual query from interventional data in a source domain, can we transport that data to answer the question in the target domain?
   • If source domain does not differ in any way from target domain, the answer is Yes.
   • If source domain differs in every way from the target domain, the answer is No
   • If the source domain differs in some ways (which we denote with a transportability node), and not in others, we use the counterfactual transportability algorithm to determine the answer
3. We provide an example where you can transport the counterfactual query from a source domain that differs in some ways from the target domain
4. We also provide an example where you cannot transport the counterfactual query because of the location of the transportability node.