Investigate ways to using io uring in read path

We currently only uses io uring in batch eviction: https://github.com/XiangpengHao/liquid-cache/blob/7c6603538321e0d6c0d22ecf76ff1f0b9a9f2229/src/storage/src/cache/core.rs#L423-L432

But the read path is still using the pread/write: https://github.com/XiangpengHao/liquid-cache/blob/7c6603538321e0d6c0d22ecf76ff1f0b9a9f2229/src/parquet/src/cache/io.rs#L146-L171

Concrete items to work on:

• understand the codebase, and write down a section (3 paragraph) to summarize how we perform io in both read and batch eviction path.
• investigate ways to use io uring for the read path. Start with easiest and simplest implementation (i.e., not the most performant)

I think @proteetpaul plan to work on this

Start by reading src/storage/src/cache/core.rs around lines 423-432 and src/parquet/src/cache/io.rs around lines 146-171 to compare batch eviction and read-path I/O. Document the current read and eviction flows in three paragraphs, then investigate the simplest way to use io_uring for reads. Done means the summary is written and an initial implementation approach is identified.