Tests for `get_labels_parameters` and `get_labels_data`

[psteinb]

a minor addition to fix the interface of sbibm. In the course of this, I saw that the task API is somewhat inconsistent with respect to naming. get_observation versus get_labels_data get_true_parameters versus get_labels_parameters

[jan-matthis]

Great, thanks for the new test case!

Here is my rationale behind the naming scheme:

• A simulator maps parameters to data, many papers use θ and x for this
• For inference, we have an observation (available through get_observation, often denoted by x_o, where the "sub o" stands for observed), which was generated by some underlying true parameters (get_true_parameters, often denoted by θ^*)
• Eventually, we obtain posterior samples, and if we simulate for those, the posterior predictive
• I wanted to have the possibility to conveniently generate labels to make plots for both of these, for which I added get_labels_parameters and get_labels_data