Have munging pipeline automatically re-read projects.csv each iteration

[jchodera]

This will avoid the need for users to restart the pipeline after updating projects.csv.

Note that we need to handle the case where the file is being edited while reading occurs. In that case, a non-compliant file should be ignored and the previous good version used. Some way to signal this status to the user (maybe a status.out file?) would be useful.