Concurrent competing optimization strategies

[jmitchell]

#3 is expected to reduce the time it takes to find the requested number of solutions, and seems like the best initial approach for adding concurrency. However, there may be other optimization opportunities, some better suited for certain problems than others. Compromising on the no-rework objective could yield solutions even faster.

Each implementation could work independently and concurrently on the same problem, streaming solutions as they're found to the coordinating agent, which in turn merges them into a single solution stream for the Backtrex client. Until a single set of optimizations proves effective across all problems, and with sufficient computing power, this could yield faster results. Additionally, tracking metrics on the performance of each implementation could help identify which are the best.

[jmitchell]

Finish #12 (Generate profiling reports) first.