test: Add secret detection benchmark dataset and ground truth

Add comprehensive benchmark dataset with 32 documented secrets for testing
secret detection workflows (gitleaks, trufflehog, llm_secret_detection).

- Add test_projects/secret_detection_benchmark/ with 19 test files
- Add ground truth JSON with precise line-by-line secret mappings
- Update .gitignore with exceptions for benchmark files (not real secrets)

Dataset breakdown:
- 12 Easy secrets (standard patterns)
- 10 Medium secrets (obfuscated)
- 10 Hard secrets (well hidden)

test_projects/secret_detection_benchmark/src/app.py

@@ -0,0 +1,19 @@
+"""
+Main application entry point
+"""
+import os
+
+# EASY SECRET #5: JWT Secret
+JWT_SECRET_KEY = "my-super-secret-jwt-key-do-not-share-2024"
+
+def init_app():
+    """Initialize the application"""
+    app_config = {
+        "name": "SecretDetectionBenchmark",
+        "version": "1.0.0"
+    }
+    return app_config
+
+if __name__ == "__main__":
+    print("Application starting...")
+    init_app()