Add focused architecture scenes, explainability, journeys, and capability-aware lenses

[Tranz007]

What this adds

This turns the full graph into a more useful Explore → Understand → Explain workflow without changing the underlying evidence model.

• focused architecture Scenes derived from meaningful graph entities, with diversified suggestions and local saved views
• bounded multi-depth Trace plus evidence-backed path finding
• deterministic “Why are these connected?” connection evidence
• hypothetical reverse-dependency blast radius kept separate from Git impact and runtime health
• local Journey assembly/playback and graph-canvas recording, preferring MP4 with a truthful WebM fallback
• Health hidden when no health evidence/adapter backs the scan
• Drift hidden until a real previous-vs-current graph comparison exists
• explicit capability metadata for Health and Drift
• product/UX decisions, roadmap, health docs, workflow contract, changelog, and README updates

Evidence boundary

Static Flow remains illustrative directionality, not realtime traffic. Journeys and focused projections only use relationships already present in the scanned graph. No hosted service, LLM, source upload, or transcoding backend is introduced.

Validation

Current-head push and pull-request CI both pass the full suite: typecheck, tests, production build, compiled CLI help, package dry-run/install smoke test, self-scan, headless viewer render, detected System Map render, and visible Flow overlay smoke test.