* dspace performs I/O for x_vals and r_vals during writeGP routine instead
of inside the driver routine.

* benchmarks.py isolates multiscale simulations as objects to avoid
  conficts inside the driver routine.

* Driver routine for optimizer is now setup independent in terms of
  objective function defination which can be specified within the
  get_x() method for the benchmark object.

* Optimizater iterations are saved in a checkpoint.pkl file for
  resurrection and post processing

* Partial dependence and convergence plots can be generated from driver
  script by passing argument visual.

Mohib