Add focused integration coverage for AI, graph, and fingerprint wiring paths

[Copilot]

This PR adds focused integration-style tests for the V2 wiring introduced across AI payload/reporting, SSRF graph traversal, and service fingerprint enrichment. The goal is to lock in cross-module behavior at integration boundaries, especially fallback and escalation paths.

• Payload generation + AI fallback

  • Added internal/payloads/generator_ai_test.go.
  • Covers WAF-detected + failed-initial-payload context with missing Anthropic credentials.
  • Asserts generator behavior remains safe and deterministic (no ai_mutation payloads emitted when AI is unavailable).

• Chain enrichment from service fingerprinting

  • Added internal/chain/portmap_fingerprint_test.go.
  • Exercises MapInternalNetworkWithFingerprint with a synthetic open Redis target.
  • Verifies service enrichment and severity upgrade fields are applied on fingerprint-confirmed exposure.

• Graph build/traversal attack-path wiring

  • Added internal/graph/wiring_test.go.
  • Builds a scored finding containing internal reachability + RCE chain signal.
  • Verifies BuildGraph emits expected graph structure and FindAttackPaths returns a high-impact RCE path.

• Report rendering AI fallback

  • Added internal/report/platforms_ai_fallback_test.go.
  • Ensures platform report rendering falls back to static template output when AI triage generation cannot run.

• Small maintainability cleanups in touched code

  • Clarified comments around adaptive payload attempt limits and credential-theft CVSS constant intent.
  • Reduced snippet-extraction duplication in fetch analysis via helper usage.

m := MapInternalNetworkWithFingerprint(ssrfProbe, fingerprintProbe)
// verifies open 6379 exposure is fingerprinted as redis and marked upgrade-critical