Revise benchmark evaluation metrics and runners

[bihius]

Summary

• Remove universal TPR/FPR targets from the evaluation plan and keep only RPS degradation as a soft lab guardrail
• Add a tagged labeled corpus runner for defensible TP/FN/TN/FP counting
• Rework go-ftw, ZAP, and Nuclei reporting so only the corpus publishes clean classification metrics
• Update collector logic, lab wiring, and docs to match the revised methodology

Testing

• Added and passed unit tests for the new metrics helpers and FTW classification logic
• Verified shell syntax for the lab runners and setup scripts
• Verified the new metrics module compiles and the benchmark Make targets expand correctly