refactor(rlm): delete dead run_rlm_turn entry points (#6511)

[Hmbown]

Refs #6511

Removes the last dead RLM entry points from #6511. The wall-clock bound is a separate question waiting on a founder decision, so this PR leaves it alone. turn_timeout() and the module doc about it are unchanged.

What changed

• Deleted pub async fn run_rlm_turn and run_rlm_turn_with_root (crates/tui/src/rlm/turn.rs) and their re-export in rlm/mod.rs. rg over the workspace and tests found no callers. The only live path is bridge.rs → run_rlm_turn_inner_with_usage.
• Their pub signatures put CodewhaleClient and Event into the crate's public API. That made rustc treat a set of test-only items as live. Removing the entry points unmasked 15 dead-code errors. This PR resolves them without changing production behavior:
  • Now #[cfg(test)] (only tests use them): run_rlm_turn_inner, CodewhaleClient::refresh_catalog_cache and publish_provider_lake_scope, ErrorEnvelope::{fatal, network}, redact_api_key_from_message, LlmError::from_http_response_with_request_context, provider_lake::replace_provider_live_snapshot, ProviderAvailability::label, ToolInspectionSnapshot::from_prepared_request, UserInputRequest::{from_value, validate}. Production already calls the *_with_limits forms, so validation at the trust boundary is unchanged.
  • Deleted (nothing used them, tests included): ToolVisibility::{RegistryOnly, Hidden}, AuthenticationErrorContext::new, AuthenticationErrorDetail::message, ErrorEnvelope::tool.
  • Event::ElevationRequired now has cfg_attr(not(test), expect(dead_code)). The TUI, runtime threads, exec agent and protocol parity all handle it, but the engine never emits it (git log -S shows no emitter in core/). Choosing between wiring an emitter and removing the sandbox-elevation flow is out of scope here. This adds one expect to the dead-code count, which goes from 259 to 260 and stays under the budget of 272.

Verification (local)

• CARGO_BUILD_JOBS=4 scripts/dev-cargo.sh clippy -p codewhale-tui --all-targets --locked -- -D warnings -A clippy::uninlined_format_args -A clippy::too_many_arguments -A clippy::unnecessary_map_or: clean. The same command without --all-targets on origin/main was also clean, and the errors listed above appeared only after the deletion.
• scripts/dev-cargo.sh test -p codewhale-tui --lib --locked -- rlm:: tool_inspection:: tools::user_input:: error_taxonomy:: llm_client:: refresh_catalog_cache provider_lake::: 201 passed, 0 failed.
• cargo fmt --all -- --check, check-blocking-calls-budget.py, check-dead-code-budget.py, check-command-crate-boundaries.py, split/module_graph.py --check: all pass.
• I did not run the full workspace suite locally. CI runs it.

🤖 Generated with Claude Code

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.