Investigating Efficiently Extending Transformers for Long Input Summarization