Max inner iterations for MMA

[smartalecH]

It would be nice to pass a maximum number of "inner" iterations for the MMA algorithm.

This is especially useful for topology optimization problems.

[stevengj]

Would also be good to be able to specify a different tolerance for the dual optimizer.

[smartalecH]

Tolerance code:

nlopt/src/api/optimize.c

Line 646 in 29e7630

nlopt_set_ftol_rel(dual_opt, LO(ftol_rel, 1e-14));

[stevengj]

Note that you can currently change this tolerance by setting the ftol_rel of the local optimizer.