feat(obd_ai): add deterministic basic health scan workflow

[virtuscyber]

Refs #4

Stacked on top of #3.

Includes:

• deterministic basic_health_scan() workflow
• structured health snapshot with findings and anomalies
• graceful handling of partially supported vehicles

[augmentcode]

🤖 Augment PR Summary

Summary: This PR introduces a deterministic, read-only basic_health_scan() workflow that produces a structured “health snapshot” plus findings/anomalies and recommended next steps.

Changes:

• Extends the approved command catalog with fuel-trim and MAF PIDs used by the scan.
• Adds protocol ID/name fields to connection metadata serialization.
• Enhances OBDAISession.connection_metadata() to safely expose protocol info when supported by the adapter/driver.
• Implements OBDAIReadOnlyToolSurface.basic_health_scan() to collect status, DTC snapshots, and a fixed set of live metrics.
• Adds analysis helpers to turn raw reads into human-oriented findings/anomalies/next steps and an overall summary.
• Adds/updates unit tests covering the new catalog entries, protocol metadata, and scan behavior for nominal and fault/partial-support scenarios.

Technical Notes: The scan is designed to be resilient to unsupported PIDs and transient read failures by returning per-command results and deriving a summarized status.

_{🤖 Was this summary useful? React with 👍 or 👎}

[augmentcode]

Review completed. 2 suggestions posted.

Comment augment review to trigger a new review at any time.

[augmentcode]

status_unavailable is labeled with severity: "warning" but it’s not added to anomalies, so summary.overall_status can remain "ok" even when monitor status couldn’t be read. Consider whether this warning should influence the overall scan status similarly to other warning-level items.

Severity: medium

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}

[augmentcode]

When a DTC read fails (ok: false), _extract_value_payload() returns None and _extract_dtc_codes() returns [], which is indistinguishable from a successful “no codes” result; this can make summary.stored_dtc_count/pending_dtc_count misleading. Consider whether the analysis/summary should preserve an explicit “DTC read failed/unknown” state.

Severity: medium

_{🤖 Was this useful? React with 👍 or 👎, or 🚀 if it prevented an incident/outage.}