Do we need exampleDistributions and exampleProblems?

[ben18785]

The former seems provide functions that would only be useful for the latter. I'd suggest that we combine these...

[ben18785]

Instead, we have a plan to refactor toy problems into dictionaries that have the same arguments that correspond to integrate as described in #26. One issue with this is how to allow toy problems to allow a variable number of dimensions. One way is to have this controlled by the length of the vector passed to the function. E.g.

• normal_any_d([0.5]) would assume that the problem was 1D
• normal_any_d([0.5, -3.2]) would assume that the problem was 2D
• normal_any_d([0.5, -3.2, 0.2]) would assume that the problem was 3D

and so on.

This seems a bit unusual to me but I guess in principle if could work?