Thanks for the report! I originally had Read and Seek but preferred to only require Read. Ideally we'd use some form of specialization for using seek where we can, but unfortunately that's not easily done right now.
In our project we have a use case where we need to read the metadata of a tar archive with large file entries (>10 GB). Adding support for Seek would allow us to reduce the runtime from minutes down to milliseconds.
I've added a PR with a proposal that adds support for readers that implement Seek while maintaining compatibility with readers that don't. Would you consider a change like this? Any alternative proposals?