API change: make __init__ of Optimizer contain the Domain and DataGenerator

[mpvanderschelling]

Optimization needs 2 things: the parameters space and the loss function.

Current state

Now, the update_step requires the DataGeneration. It would be cleaner if this data generator is stored as a self.data_generator argument during creation of the Optimizer

Requested change

These two arguments should be required during creation of the object.
Also then the _construct_model method can collapse into the __init__ or __post_init__ if we are using dataclasses again

API for Optimization

There are three ways:

• Using built-in defaults by importing a function

from f3dasm_optimize import cmaes
experiment_data.optimize(optimizer=cmaes(population=30, seed=123), data_generator = ..., ...)

• Using built-in default by string argument

experiment_data.optimize(optimizer='cmaes', data_generator = ..., hyperparameters={'population': 30, 'seed':123})

• Custom optimizer

class MyOptimizer(Optimizer):
    def update_step(self):
           data_generator = self.datagenerator
           data = self.data
           return ExperimentData # new samples

experiment_data.optimize(optimizer=MyOptimizer, data_generator= ... , hyperparameters({'arg1': arg1, 'arg2': arg2})