Polaris step "complete" for certain types of failures

[cbegeman]

It seems desirable for me for Polaris steps to complete when they run correctly, even if the step fails. An example of when this comes up is when the convergence rate is computed to be less than the threshold, in which case polaris_step_complete.log is not generated.

[xylar]

I agree, that makes sense. I shouldn't be hard to implement.