Add prominent warnings for protection score proxy metric

[Claude]

The protection score is a proxy metric derived from ResNet18 feature drift, not validated against real-world mimicry attacks like DreamBooth or LoRA. Users were not adequately warned that high scores don't guarantee actual protection.

Changes

Core Metrics (src/auralock/core/metrics.py)

• Added warning constants: PROTECTION_SCORE_WARNING, PROTECTION_SCORE_DISCLAIMER, PROTECTION_SCORE_METRIC_TYPE
• Modified get_protection_readability_report() to include warning metadata in all responses
• Updated assessment labels to include "(proxy space)" suffix (e.g., "Strong (proxy space)" instead of "Strong")

CLI Output (src/auralock/cli.py)

• Added warning row to protection readability table
• Display prominent yellow warning after protection reports in protect, analyze, batch, and benchmark commands
• Warnings appear both inline in tables and as standalone messages for visibility

Example output:

Protection score: 48.2/100
Assessment: Strong (proxy space)
⚠️  Warning: Protection Score is a proxy metric. Real-world effectiveness is NOT validated.

⚠️  Protection Score is a proxy metric. Real-world effectiveness is NOT validated.
This score measures drift in ResNet18 feature space, NOT actual mimicry prevention...

Documentation (README.md)

• Moved warning from buried text to prominent blockquote before benchmark results table
• Emphasizes lack of empirical validation against real-world attacks

Tests (src/tests/test_metrics.py)

• Added 3 test cases verifying warning metadata presence
• Verify assessment labels include "(proxy space)" suffix
• All existing tests pass with new changes