We need to decide (and implement) some additional heuristics for when we're choosing an index as the rhs of an index probe during INLJ plan enumeration. For e.g. we don't take into account index uniqueness