//! Positional file I/O.
//!
//! The engine never seeks. Every read and write carries its own absolute
//! offset (`pread`/`pwrite`), which is what allows N workers to touch one file
//! concurrently without a lock and without a shared cursor. It is also why a
//! 100 GB file costs the same resident memory as a 100 MB one: nothing is ever
//! buffered beyond the chunks currently in flight.
pub use ChunkReader;
pub use ;