The Unreasonable Effectiveness of Randomized Representations in Online Continual Graph Learning