Use latest doublets-rs (0.3.0) and switch to stable Rust

[konard]

Summary

Resolves the two core requirements of #23:

1. Upgrade to the latest doublets-rs — replaced the git-pinned pre-release
   0.1.0-pre+beta.15 with the published crates.io release doublets = "0.3.0"
   (the newest version; confirmed via cargo search doublets).
2. Switch to stable Rust — rust-toolchain.toml and the CI workflow now use
   stable; all nightly feature gates (allocator_api, generic_associated_types)
   and nightly-only API usage have been removed/migrated.

It also delivers the documentation deliverables the issue asks for: a deep case
study in docs/case-studies/issue-23/
covering every requirement (R1–R8), online research, root-cause analysis,
per-requirement solution plans, and a survey of existing components/libraries.

API migration map (nightly beta.15 → stable 0.3.0)

Old (nightly)	New (stable)
#![feature(allocator_api, generic_associated_types)]	(removed — GATs stable since 1.65; allocator_api provided via allocator-api2)
Alloc<T, std::alloc::Global>	doublets::mem::Global<T> (built with Global::new())
doublets::data::LinkType	doublets::data::LinkReference
T::ZERO	T::from_byte(0)
value.as_i64()	as_i64(value) helper wrapping TryInto<i64>
handler(link)? (relied on nightly Try for Flow)	if handler(link).is_break() { return Flow::Break; }
doublets::parts::LinkPart	doublets::unit::LinkPart
toolchain nightly-2022-08-22	toolchain stable

The PostgreSQL backend logic (client.rs, transaction.rs) needed only
mechanical symbol renames because doublets's Links<T>/Doublets<T> traits
kept their Link<T>-based handler signatures. Exclusive<T> already satisfies
the new Send + Sync bound on Links<T>.

Verification

$ cargo build --release --all-features --manifest-path rust/Cargo.toml   # OK
$ cargo bench --no-run --all-features                                    # OK (all bench binaries compile)
$ grep -rn '#![feature' rust/src rust/benches                            # no matches

rustc 1.95.0 (stable). Benchmarks connect to PostgreSQL only at runtime, which
CI provides via its postgres service; locally, compilation is the verifiable
step.

Files changed

• rust-toolchain.toml, .github/workflows/rust.yml — toolchain → stable
• rust/Cargo.toml, rust/Cargo.lock — doublets = "0.3.0"
• rust/src/{lib,client,transaction}.rs, rust/src/benched/* — API migration
• rust/benches/** — removed #![feature], Alloc<…> → Global<…>
• docs/case-studies/issue-23/README.md — case study (R3–R7)

Closes #23

[konard]

Working session summary

CI is still running the benchmarks. The background poller (bd41k7kyn) will notify me when the run completes. Let me summarize the current state while waiting.

All implementation work is complete and pushed. Here's where things stand:

Completed:

• ✅ R1 — Latest doublets-rs: doublets = "0.3.0" (crates.io, confirmed latest)
• ✅ R2 — Stable Rust: rust-toolchain.toml + CI use stable; all #![feature(...)] gates and nightly-only API removed. Verified locally with cargo build --release and cargo bench --no-run on rustc 1.95.0.
• ✅ R3–R7 — Case study: docs/case-studies/issue-23/README.md with requirements table, online research, root-cause analysis, per-requirement solution plans, and component survey.
• ✅ R8 — Single PR: commits on issue-23-dff35da4182d; PR Use latest doublets-rs (0.3.0) and switch to stable Rust #24 retitled (removed [WIP]), description rewritten, marked ready for review.

In progress: Waiting on CI run 26664990689 (the "Benchmark Rust version" workflow, which builds on stable and runs cargo bench against a live PostgreSQL service) to confirm green.

I'll report the CI conclusion as soon as the run finishes.

---

This summary was automatically extracted from the AI working session output.

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost estimation:

• Public pricing estimate: $8.198766
• Calculated by Anthropic: $8.165360
• Difference: $-0.033406 (-0.41%)

📊 Context and tokens usage:

Claude Opus 4.8: (2 sub-sessions)

1. 116.3K / 1M (12%) input tokens, 28.5K / 128K (22%) output tokens
2. 97.2K / 1M (10%) input tokens, 28.8K / 128K (23%) output tokens

Total: (21.2K new + 343.8K cache writes + 9.0M cache reads) input tokens, 57.3K output tokens, $8.198766 cost

🤖 Models used:

• Tool: Anthropic Claude Code
• Requested: opus
• Model: Claude Opus 4.8 (claude-opus-4-8)

📎 Log file uploaded as Gist (3362KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.

[konard]

✅ Ready to merge

This pull request is now ready to be merged:

• All CI checks have passed
• No merge conflicts
• No pending changes

---

Monitored by hive-mind with --auto-restart-until-mergeable flag