Data-Driven Learning of Large Causal Structures “QuICly” using Quotient Graphical Independence Models