Parset files are now collected from stdout.

This is a small MR that slightly modifies how parset files are created. Rather than copying a temporary file that is generated by CWL at runtime, it prints the contents of this file to stdout. stdout is then captured by CWL.

I feel this approach is slightly cleaner, because it removes some overhead (the setup of the arguments), creates an output file with sensible permissions (rather than read only for the user), and is perhaps a bit more maintainable (as this approach closely follows what is described in the CWL user guide.