(Python) functions as non-linear constraints in DoE

[dlinzner-bcs]

Multiple use case owners have their constraints in terms of complicated functions. Sometimes they have them as (python) functions already available. Since we already use automatic differentiation, we could support the non-symbolic functions as non-linear constraints.

[Osburg]

@dlinzner-bcs sounds cool :) How would it be intended to map the variables names to the function inputs then? By requiring the function to have input names corresponding to the variable names and the inputs are given as a dict {"var1": values_for_var1, "var2": values_for_var2...} which we can generate from the model matrix?