Define a structural causal model

The structural causal model used in the publication is printed below.
The setup is easy:

1. Define the background noise variables (i.e. variables not caused by any other variable)

2. variable is the internal name

3. label is a longer name; when these corerspond to measurements of the images (like here size and variance), they will be used to sample images
4. type noise vs dependent
5. distribution where to draw the variable from; for the dependent variables, this is the conditional distribution

6. param_1 and param_2 are the canonical parameters for the distribution (e.g. location and scale for Normal)

7. Define the relationships between the noise variables and the dependent variables, using b_... columns to define coefficients from the noise variable to ... in a linear model.