Codex: add agent-readable report intelligence

[hardcoreerik]

Summary

• Adds report_intel.py with read-only intelligence helpers for latest-run summaries, strongest unknown APs, latest-run comparison, suspicious handshakes, evidence health, and planned integration surfaces.
• Extends wardrive_service.py and wardrive_cli.py so Claude/Codex can call these as JSON/stdio actions.
• Documents the AI bridge commands and adds a roadmap for import adapters, WiGLE enrichment, packet-level filters, survey maps/floorplans, report quality, and a future MCP wrapper.

Verification

• python -m py_compile report_intel.py wardrive_service.py wardrive_cli.py
• python wardrive_cli.py summarize-latest-run F:\WardriveAnalytics --compact
• python wardrive_cli.py compare-latest-runs F:\WardriveAnalytics --compact
• python wardrive_cli.py strongest-unknown-aps F:\WardriveAnalytics --limit 5 --compact
• python wardrive_cli.py suspicious-handshakes F:\WardriveAnalytics --limit 5 --compact
• python wardrive_cli.py evidence-health F:\WardriveAnalytics --compact
• stdio bridge request for summarize_latest_run against F:\WardriveAnalytics

Collaboration note

Claude can build the report-generation UI/report prose on top of these stable JSON actions without scraping GUI logs or raw report HTML. The comparison output is intentionally capped to avoid flooding agent context when two runs differ by tens of thousands of BSSIDs.