On Ranking-based Tests of Independence