Imaginary powers when running with optimal stiffness/damping for each sea state

[rebeccamccabe]

Graph generated from @MadisonDietrich 's commit c549142 using run_single.m. Left image is with normal non-optimal stiffness/damping, and right is with the new optimal stiffness/damping. The optimal setting generates about twice as much average power but there are still some sea states (like the one highlighted in the figure) which actually produce less power in the optimal setting.

[rebeccamccabe]

I don't get any errors when running gradient_optim.m. When running test cases, it encounters the imaginary power after it attempts a wider tolerance on the multiplier (up to 1.001, see line 19 of pick_which_root.m), during the parameter sweep.