fix: fail closed benchmark guard consumption

[Pigbibi]

Summary

• preserve zero risk, leverage, and risk-asset scalars when the mounted benchmark guard is blocked
• mark defense required so a separately approved strategy consumer can move to its defensive sleeve instead of retaining risk on unavailable/stale benchmark data
• retain the plugin's shadow-only/no-order execution controls

Validation

• uvx ruff check src/quant_strategy_plugins/market_regime_control_plugin.py tests/test_benchmark_drawdown_guard.py
• uv run --extra test pytest -q tests/test_benchmark_drawdown_guard.py tests/test_market_regime_control_plugin.py tests/test_strategy_plugin_runner.py tests/test_plugin_policies.py tests/test_plugin_catalog.py