fix: release bounded telemetry claims on completion

[Joncallim]

Summary

• Own each bounded telemetry stats attempt through its timeout and completion.
• Prevent detached task cancellation from leaving the two-slot cache budget occupied.
• Add a Unix-socket regression for a 200 response with a non-completing body.

Verification

• cargo fmt --manifest-path crates/Cargo.toml --all --check
• npm run test:rust:daemon (177 passed)
• npm run test:rust:core (69 passed)

No contract files changed. Contract tests were not run in this isolated worktree because its dependencies are absent. Rust lint was attempted but the host ran out of disk space while creating a duplicate worktree target.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-02T17:42:16.731472Z	08a81c3	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 08a81c3721

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Keep telemetry waits off the snapshot refresh loop

When either stats request is slow or reaches the 750 ms timeout, awaiting the whole join_all blocks refresh_cache; because refresh_loop starts its two-second sleep only after that call returns, every affected tick stretches the Docker snapshot cadence to roughly 2.75 seconds plus collection time. This contradicts the two-second cadence relied on in docs/architecture/PROVIDER_SCHEDULING_AND_MODEL_REVISIONS.md and makes all topology and health views staler whenever the stats gateway stalls; retain claim ownership without putting the telemetry timeout on the snapshot scheduler's critical path.

Useful? React with 👍 / 👎.