Don't require original source data for pipes that don't use it
Currently, one way to run CTAP is to build a Cfg structure that points to a set of raw files (in any format) on disk and starts processing them as specified. Step sets save new copies of the data to disk, using EEGLAB *.set format
PROBLEM: if you rereun CTAP with a subset of the pipe that starts at some stepset > 1, then CTAP does not NEED to have access to the original data, and yet it still forces us to give access, since the Cfg structure will fail if the defined data is not there. That's a drawback on distributed setups, especially for larger datasets, where you might want to have original data on 1 disk and write CTAP output to another disk.