Starting values for reform runs

[jdebacker]

Currently, SS.runSS will use the SS solution from the baseline run for the initial guess at the reform run (so long as the size of the state space is the same). However, if the reform policy is substantially different than the baseline, this means the starting values aren't very good. But OG-Core provides no way to override using the baseline solutions as starting values.

A few ideas so solve this issue:

1. Add robustness to initial values by trying new starting values if a solution isn't found (as proposed in Issue #776).
2. Create a parameter, use_baseline_solution that applied to the reform run. If this is false, then the reform run will not use the baseline solution for its initial guesses.
3. Start with the baseline solution as initial value and if an eq'm fails, then try the default (or user provided) starting values.