On the Validity of Head Motion Patterns as Generalisable Depression Biomarkers