Applying Linearly Scalable Transformers to Model Longer Protein Sequences