Aequa: Fair Model Rewards in Collaborative Learning via Slimmable Networks