TraKDis: A Transformer-based Knowledge Distillation Approach for Visual Reinforcement Learning