Add method to transform linear equality constraints to variable dependencies

[schmoelder]

Linear equality constraints can always be transformed to a variable dependency which has several advantages:

• It reduces the number of variables
• Many optimizers are not well suited for equality constraints (especially population based algorithms)

We could simply do this transformation "under the hood".

Here, we might want to discuss how to generally deal with transformed OptimizationProblems since adding different transforms (variable normalization, conditioning, equality normalization) could make the interface a bit too cluttered (e.g. upper_bounds, upper_bounds_transformed, upper_bounds_independent_transformed, etc.).

Options:

• Add getter methods with flags to get the values (e.g. get_bounds(normalize=True, independent=False)
• Create "child" OptimizationProblems: opt_new = transform_problem(opt)