Io_uring Without Readahead

Turso implemented readahead for its io_uring backend to improve performance by submitting multiple read requests concurrently instead of waiting for each one sequentially. Testing with TPC-H benchmarks showed that readahead reduces device requests from ~196,000 to ~16,300 by allowing the kernel to merge adjacent I/O operations, resulting in larger average read sizes (56.53 KiB vs 4.37 KiB) and better overall throughput despite fetching more total pages.