fix(ci): fix linting issues

[TaylorMutch]

Summary

Fix the lint failures currently present at the tip of main.

Related Issue

N/A

Changes

• Restore the gateway shutdown sender name in openshell-server so shutdown_tx.send(true) compiles.
• Allow the GCE metadata emulator modules to be dead code on non-Linux builds, where their production caller is compiled out.
• Gate the sandbox tokio::time::timeout import to Linux for the same reason.

Findings

main currently fails mise run pre-commit because of two independent regressions.

First bad commit for the sandbox lint failure:

• 48545cfb feat(sandbox): add GCE metadata emulator for Google Cloud (#1763)
• Verified by running mise run rust:lint at 48a7d09e, which passed, then at 48545cfb, which failed.
• Cause: PR branch checks run Rust jobs only on Linux runners. The GCE metadata emulator is used only behind #[cfg(target_os = "linux")], so Linux clippy sees the modules as used. On macOS/non-Linux, the production caller is compiled out and -D warnings turns the modules/import into dead-code and unused-import errors.

Second regression at the current main tip:

• 4ee27d99 feat(sandbox,providers): add aws-bedrock as a recognized inference provider (#1704)
• Cause: the PR renamed the first gateway shutdown sender to _shutdown_tx, but run_server still calls shutdown_tx.send(true) later. The PR passed before merge because branch protection has strict: false; it was not rebuilt against the current main merge result after 4c75b854 fix(server): share gateway shutdown channel (#1945) changed the shutdown-channel shape.

CI coverage that would have caught this before merge:

• A required macOS/non-Linux Rust lint job running mise run rust:lint or cargo clippy --workspace --all-targets -- -D warnings.
• Strict required checks or merge queue testing so PRs are validated against the current main merge result before merge.

Testing

• mise run rust:lint passes
• mise run pre-commit passes
• git diff --check passes
• E2E tests added/updated (not applicable)

Checklist

• Follows Conventional Commits
• Commits are signed off (DCO)

[elezar]

Another one for #1946?

[TaylorMutch]

Another one for #1946?

Yes!