examples: add Agent Payment Guard HTTP service

[safal207]

Motivation

• Expose the existing Agent Payment Guard as a local HTTP authorization boundary that external AI agents/adapters can call before execution.
• Reuse existing guard decision and audit logic to avoid duplicating core rules while providing a simple service surface for testing and integration.

Description

• Add examples/agent-payment-guard/payment_guard_service.py implementing a small ThreadingHTTPServer with endpoints GET /v1/health, POST /v1/payment-proposals/evaluate, and GET /v1/audit/records that reuse decide, append_audit, and get_previous_hash from payment_guard.py.
• Implement enforce and shadow mode semantics and return execution_allowed, would_block, and audit_hash in evaluation responses.
• Read and return recent audit records from .proofpath/audit.jsonl with a bounded limit query parameter.
• Add examples/agent-payment-guard/run_service_check.sh to exercise the service end-to-end (health, enforce + shadow evaluation, audit retrieval, and hash-chain verification).
• Add docs/agent-payment-guard-service.md with startup instructions and copy-pasteable curl examples.

Testing

• Ran bash examples/agent-payment-guard/run_demo_check.sh and it passed (demonstrates CLI demo still works).
• Ran bash examples/agent-payment-guard/run_service_check.sh and it passed (service health, enforce/shadow semantics, audit endpoint, and audit_hash checks succeeded).
• Ran python3 scripts/verify_audit_log.py .proofpath/audit.jsonl and it passed (audit hash-chain verification succeeded).

---

Codex Task