Adopt Criterion.rs for Benchmarks

## What is the problem your feature solves, or the need it fulfills?

I was exploring the LinkedList implementation in the pingora-lru crate and wondered whether using Criterion.rs could make your benchmarks more reliable. Right now, the linked list benchmarks (and possibly those in other crates as well) use some ad-hoc timing that can feel noisy and hard to reproduce, especially when iterating over lists. Adopting Criterion would give built‑in warm‑up, statistical analysis, and clear “before vs. after” reports.

## Describe the solution you'd like

I’d like to suggest migrating the existing benchmarks in pingora-lru to use Criterion.rs. If this seems like a valuable change, I’d be happy to contribute the initial implementation myself!

## Describe alternatives you've considered

- [iai-callgrind](https://github.com/iai-callgrind/iai-callgrind), another benchmarking library that uses CPU cycle counting instead of timing. It’s a great tool, but Criterion provides statistically driven benchmarks with confidence intervals, which I’m more familiar with.

## Additional context

- [Criterion book](https://bheisler.github.io/criterion.rs/book/criterion_rs.html)
- [Criterion repository](https://github.com/bheisler/criterion.rs)

I’m suggesting Criterion specifically because I’ve used it before and find its workflow effective, but I think both Criterion and iai-callgrind are strong options depending on the goals.