SymbolicNLP --> NlpBuilder

[jaeandersson]

The SymbolicNLP class was written to allow import of NLPs in the .NL format as SX expressions. I would like to improve this functionality and extend extend the scope of the class to allow step-by-step construction of NLPs similar to the way DaeBuilder (previously SymbolicOCP) now allows step-by-step constructions of DAEs, cf. #1358. This class would also be used for constructing LPs and QPs, just like the .NL format is not only used for NLPs.

The problem formulation could include different types of constraints and objective function terms, such as least-squares objective terms, inequality and equality constraints separately, linear constraints separately. Like for DaeBuilder, there should be support for reformulating the problem into a suitable form.

Tasks:

• MX instead of SX
• Support for intermediate variables
• Support for least-squares terms
• Expose functions for adding optimization variables, parameters, objective terms and constraint terms

Cf. #755