Better checkpointing

I have to deal with some interrupted runs, and when re-running I-TASSEr, it seems that it doesn't always restart from the right place.

For example, if a run is interrupted after model1.pdb is produced (say it is working on model2.pdb), restarting the run skips the model generation step. There seem to be similar issues for the COACH/COFACTOR steps afterwards. I'd suggest that you do something like "touch .model-generation.done" when that step is complete and look for this file to determine whether the step has been completed.