Add prompt injection false positive tests

[lleonardo-franco]

Summary

• Add negative PromptInjectionGuardrail coverage for benign inputs that contain suspicious words such as ignore, system, and prompt injection.
• Cover code comments, prompt engineering documentation, user formatting instructions, technical system logs, and version-specific guidance.
• Assert each benign case passes with no detections.

Validation

• npm test -- src/guardrails/__tests__/built-in-guardrails.test.ts --runInBand
• npm run build:tsc
• git diff --check

Resolves agentguard-ai/tealtiger#85