Use `blockio-uring` with support for oversized batches

[jorisdral]

The lookup macro-benchmark speeds up by about 0.6 seconds

Before: macro-before.log After: macro-after.log

[dcoutts]

That's about a 6% improvement. Not bad.