[2.0] Add DuckDB E2E query optimization task

[joyemang33]

Summary

Adds a new Frontier-CS 2.0 systems task, duckdb_e2e_query_optimization, where agents patch a pinned DuckDB checkout to improve end-to-end TPC-H style analytical query performance while preserving correctness.

The task includes Harbor-specific agent/judge Docker images, an async patch-submission workflow, a deterministic evaluator with hidden scale-factor groups and randomized query order, and patch-policy guardrails for optimizer/execution-focused DuckDB changes.

Please read CONTRIBUTING.md before submitting.

Type of Change

• New research problem
• New algorithmic problem
• New Frontier-CS 2.0 problem
• Bug fix
• Documentation update
• Other:

Testing

• PYTHONPYCACHEPREFIX=/private/tmp/frontier-cs-pycache python3 -m py_compile 2.0/problems/duckdb_e2e_query_optimization/evaluator.py
• UV_CACHE_DIR=/private/tmp/frontier-cs-uv-cache uv run --no-sync frontier show 2.0 duckdb_e2e_query_optimization
• Generated the Harbor task through frontier_cs_2_0.main for duckdb_e2e_query_optimization
• Built and smoked the Docker images:
  • frontiercs/duckdb-e2e-query-optimization-agent:experimental-v1.5.3
  • frontiercs/duckdb-e2e-query-optimization-judge:experimental-v1.5.3
• Ran malicious patch-policy smoke checks for source deletion, denied-path rename/copy, TPC-H string hardcoding, denied CMake edits, and allowed optimizer subdir CMake wiring
• Ran Harbor trial: frontier-cs-2-0-duckdb-e2e-query__XNXGS9L
  • score: 2.8865579012232834
  • reward: 0.028865579012232835
  • geomean speedup: 1.0202095982916792
  • successful submissions: 3
  • used best submission: true
• Checked latest trial feedback for hidden leakage; no per_query, concrete hidden sf/q labels, correctness_mismatches, or stderr_tail were exposed

Checklist

• Code follows the project structure and conventions
• Self-review completed
• Documentation updated (if applicable)

CI Validation (for new problems)

When adding new problems, CI will automatically validate that your reference solution achieves score > 0.

• Algorithmic problems: Include reference.cpp in your problem directory