qcxms: remove 'input in' files from neutral run output for each trajectory and use single file as input, as the file for each trajectory contains the same parameters anyway

[hechth]

This would reduce the number of datasets generated quite drastically as it would result in a single dataset over a dataset for each trajectory.