...and therefore also not implemented in fortran (#52), python, etc.

The NLOPT_LD_CCSAQ algorithm allows the user to supply an approximation to the Hessian, via
the set_precond_min_objective() and set_precond_max_objective() functions. Unfortunately, it looks like these are not included in nlopt-in.hpp (at least I assume this is the problem). I'm not well-versed in writing wrappers; can someone give me a hand to implement the changes?