Delay model generation: read in existing checkpoint file if one exists

The `timing_characterization_client.py` script should both read and write a checkpoint file for the data points that it collects. If there is one that is read in, then the script will use this already-collected data and not rerun characterization for those points.

However, as currently used by top script `run_timing_characterization.py`, it does not seem to be working correctly. All characterizations points are rerun, and a previously-existing checkpoint file is overwritten.