Laying out contributions

[davidlove]

1. We expanded the classical overlapping batches method from simulation to stochastic programming
   a. Each batch now contains an optimization problem, which was not included in the original case
   b. The optimization problem effectively means that batches themselves have properties not cleanly reducible to the samples contained in batches.
   c. We extend the original asymptotic analysis to the stochastic programming setting.
2. Computational experiments show that variance reduction is achieved with relatively small sample sizes.