RLM: Simplify code

[snimu]

Description

Simplifies RLM code without touching functionality.

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Test improvement

Testing

• All existing tests pass when running uv run pytest locally.
• New tests have been added to cover the changes

Checklist

• My code follows the style guidelines of this project as outlined in AGENTS.md
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

Note

Major internal cleanup with no intended behavior changes, focusing on maintainability and consistency.

• Worker script generation: Replace large inline templates with _build_python_worker_script_template(sandboxed=...); extract reusable blocks (guardrails, FIFOs, FS context, jail); generate both local _RLM_WORKER_SCRIPT_TEMPLATE and sandbox _RLM_SANDBOX_PY_WORKER_SCRIPT_TEMPLATE from one source.
• Env var setup: New _build_worker_env_vars() used by local and sandbox executors; toggles inclusion of code-exec restrictions.
• Metrics: RLMMonitorRubric now auto-registers simple state metrics via a factory (_SIMPLE_METRICS, _make_state_metric); dynamic per-root-tool metrics retained.
• REPL flow: Factor shared logic into _call_repl() and _maybe_add_context_warning(); call_python_repl appends execution time, call_bash_repl does not; warning text parameterized.
• Tool docs/types: Introduce _append_tool_docs() to avoid duplication; annotate OAI tool lists with ChatCompletionFunctionToolParam; simplify root tool schema handling.
• Sub-LLM accounting: Use _extract_tokens_from_response helper; llm_batch summary now shows prompt vs completion tokens separately; minor log/message tweaks.

Risk is primarily around template generation and executor env var changes affecting worker startup and token/metric reporting.

^{Written by Cursor Bugbot for commit c619cc8. This will update automatically on new commits. Configure here.}