feat(risk): add long-horizon policy composer

[Pigbibi]

Summary

• add a pure, generic long-horizon risk-policy composer for strategies, combinations, and plugins
• calculate a 10%-step scale frontier from frozen net-of-cost P3 paths against an unlevered benchmark
• require walk-forward, bootstrap, and stress coverage of at least 252 sessions each; missing evidence parks
• offer three transparent human preference templates while binding all concrete advice to P1/P2/P3/plugin digests
• document that results are advisory and cannot change policies, parameters, P4/P5/P6, accounts, or orders

Verification

• ruff check python/scripts/long_horizon_risk_composer.py python/tests/test_long_horizon_risk_composer.py
• python -m unittest discover -s python/tests -q (230 passed)
• git diff --check

No strategy configuration, execution policy, cloud resource, broker connection, or order is changed.