Improving Test–Retest Variability of Visual-Evoked Responses in Multiple Sclerosis: Implications for Trial Design