Bells and whistles for attacks / nonconvex solvers

[GeoffNN]

It has been shown that various improvements on textbook optimization methods allow for notable gains in attack performance. This should be implemented either in the Attacker class or directly in the optimize_method functions.

• [ ] Random restarts
• [ ] Backtracking line-search
• [ ] Stop when adversarial example is misclassified. This could be implemented in a callback passed to optimize_method.
• [ ] use_best option for all methods