Reconquering Bell sampling on qudits: stabilizer learning and testing, quantum pseudorandomness bounds, and more