Text this: Partial end-to-end reinforcement learning for robustness towards model-mismatch in autonomous racing’