Faster IO through io_uring (2019)

I'm posting this because, while I had a reasonable idea of how io_uring works at a high level, I had never written any io_uring-using code myself, and I thought this was a very good introduction to the concrete details of its operation.