Replace once_cell dependency with std LazyLock and OnceLock

[abd0-omar]

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

once_cell provides lazy initialization primitives (Lazy, OnceCell). It's now in the standard library as LazyLock and OnceLock.
Continuing to use once_cell adds an unnecessary external dependency when the same functionality is now available natively in Rust's standard library.

Describe the solution you'd like

replace all it's usages with the standard library equivalents as it will be a 1 to 1 replacement:

• once_cell::sync::Lazy -> std::sync::LazyLock
• once_cell::sync::OnceCell -> std::sync::OnceLock

except for try_insert(), found in pingora-runtime/src/lib.rs, as it is not stable yet issue, it's alternative would be doing set() then get().unwrap() till it gets stabilized.

Describe alternatives you've considered

Keep using once_cell

[ExpertVagabond]

Before anyone starts a fifth attempt: this replacement has been written four times already, and none of it has landed.

PR	State	Opened	Files	Note
#732	open	2025-11-09	33	opened by this issue's author
#735	closed	2025-11-09	45	closed by its own author in favour of #956
#881	closed	2026-05-18	35
#956	open	2026-08-13	36

Current state of the two that are still open, as of today:

• Neither has a single review or comment. Not from a maintainer, not from anyone.
• Both now report mergeable: false, mergeable_state: dirty, so both have drifted into conflict with main while waiting.
• once_cell is still referenced in 8 workspace Cargo.toml files, so nothing has been merged by another route either.

MSRV is not the blocker. That was my first guess, and it is wrong. The README states the current MSRV is 1.85, and .github/workflows/build.yml pins its msrv matrix entry to 1.85.0. LazyLock stabilized in 1.80 and OnceLock in 1.70, so both are comfortably available. The try_insert caveat in the original description remains the one genuine exception.

So the change appears technically fine and has stalled for reasons that are not written down anywhere. That is the part worth a maintainer's thirty seconds:

• If this is wanted, one of replace dep once_cell with std library equivalent LazyLock/OnceLock #732 or Replace once_cell with standard library primitives #956 needs a rebase and a review, and saying which one would stop the duplication immediately.
• If it is not wanted, closing this issue would save the next person a day of work, and would have already saved four people theirs.

Either answer is more useful than a fifth PR, which is why this is a comment.