docs: expand Appendix D technique inventory for issue #583

[jmanico]

Summary

• add Appendix D inventory entries for concrete adversarial, privacy, supply-chain, and monitoring techniques discussed in Gap analysis: OWASP AI Exchange controls vs. AISVS - proposal to discuss adversarial ML coverage #583
• keep the changes non-normative by mapping each technique back to existing AISVS requirement IDs
• strengthen Appendix D as an implementation-oriented inventory without introducing duplicate requirements

Included techniques

• structured-data injection validation and context poisoning detection
• training-time poisoning detection examples such as RONI, influence scoring, gradient analysis, and activation clustering
• third-party model behavioral acceptance testing
• differential privacy budget exhaustion enforcement and PATE
• ensemble disagreement analysis, randomized input transformation, and confidence obfuscation
• runtime behavior drift detection examples

Issue linkage

Refs #583