Turboflow

- Change calculation of violation of inequality constraints: in pygmo (and TurboFlow), constraints are defined to be less than zero. The violation hade to be change to be np.maximum(c_ineq, 0) from np.minimum(c_ineq, 0).
- Change definition of constraints in configuration file. To define an uppar and lower bound for a constraint, two constraints must be defined in the configuration file. However, if the constraints are a dictionary, where the key is the variable name, only one of the two constraints will be registered. The constraints must be defined as a list of dictionaries, where a key is the variable of the constraint.

- Correct hub_tip_ratio_in calculation in calculate full geometry. This affect performance analysis!
- Add option to initialize optimization with feasable point. By giving the optimization a solved performance analysis problem, the result is converted to an initial guess for the design optimization.
- However, the geometry model is not fully consistent, as the stagger angle and maximum thickness is calculated for design optimization, but given for performance analysis.
- Also, the default initial guess and feasable initial guess can give very different geometries.

- Blade jet ratio implemented as design variable
- Updated prepare_geometry according to new set of design variables. Blade jet ratio assign a geometry, either hub, mean or tip, which is constant throughout the turbine.
- Updated geometrical model was somewhat detrimental to the convergence. Calculating the opening from the throat area was the source of this. The debugging was performed by creating a new branch, check_optimization, which holds the old geometrical model, and converges better.
- In addition to the step size for the finite difference approximation for the Jacobian, the convergence was sensitive to the scaling of the objective function
- Plot function to plot turbine in axial-radial plane