Documentation, Discussion of determinsttic parameter

[LukasFehring]

In the scenario class, the deterministic parameter is set to false by default. This might have some unintended consequences. However on first glance it appears that everything works as expected due to the parameters n_seeds and max_config_calls.

However, this needs to be discussed, possibly reworked and definitely documented.

[LukasFehring]

We discussed this internally. We believe that the best course of action is adapting this hyperparameter for each facade. The default should remain deterministic = False. As a result, the defaults should be adapted such that the same configuration can be executed multiple times.
As a result, the documentation and examples need to be adapted. Currently, they imply a deterministic execution.

[benjamc]

Two aspects here:

1. If the function is non-deterministic, it should be modeled as such. Therefore the GP should receive a deterministic parameter and the kernel composition should be adapted. We could also think of a user-defined kernel / noise level to be passed or the type of noise. This information is not important in the facade, but only in the model.
2. Intensification. Find out what arguments relate to non-determinism (e.g., n_seeds, max_config_calls.

Summary: For a non-deterministic function, the user needs to specify the intensification resources and optionally whether it should be modeled in the GP if used.