Allow contrasts to be set in Between Subject ANOVA

[nicholst]

In particular for Between Subject ANOVA, k groups, 1 scan per subject, we should allow other arbitrary (but valid) contrasts. For example, if someone is examining 3 groups, they could test for [-1 0 1] or [-1 1 0].

Requested by Zhijiang Wang.