When clipping, the origin of the great arc doesn’t change, so we can cache a
number of intermediate values and accelerate the clipping computation.