Thinking back to Assignment 1, one way that comes to mind for optimizing the performance for this part would be to make the plane be the bounding box of the triangle. In Assignment 1, to optimize our rasterizer we only tested pixels within the bounding box of the triangle. It makes sense to me that the plane in this case would also be the bounding box of the triangle, as the bounding box would be a plane that encompasses the triangle completely. It seems like the idea of the bounding box would be a very effective way to test for ray intersection with a triangle