generate_runscripts.py should create scripts that have a continue mode

[cwpearson]

where, if an output file exists and doesn't have "error_occurred" in it, it will not re-run that benchmark

This could be an option at the top of the generated script so it would be easy to turn off.