Pass on linear constraints to optimizers that can handle them

[ChristianZimpelmann]

Current situation

It isn't possible to implement the following three constraints (bounds):

• $$0 \leq para_1$$
• $$0 \leq para_2$$
• $$para_1 + para_2 \leq 1$$

One can introduce a third parameter and make use of a ProbabilityConstraint, but this seems a bit hacky.

Desired Situation

Some packages/optimizers allow for this set of constraints, e.g. SLSQP by scipy or DifferentialEvolutionSolver by mystic

Proposed implementation

• pass on linear constraints to optimizers that can handle them
• Alternative: improve on reparameterization within optimagic (not sure whether this is possible)