Add the Scouting Atlas presentation

[iantocristian]

A visual explanation of the semantic plane — scouting, artifacts, freshness, and agent-side retrieval — as a companion to the jscout Atlas (#61). Seven questions in a newcomer's reading order, each answered with one inline-SVG figure.

1. What the scout produces — the five families as a dependency picture: repository classification gates planning; workflows and cards read hashed source; summaries and concepts read only other artifacts
2. One scout run — plan → prepare → execute → publish, with the fingerprint computed before the ledger claim and the scout_runs row written before the model call
3. Artifact anatomy — semantic_artifacts → body_json {text, evidence[]} claims → semantic_supports → hashed line ranges; relations and embeddings
4. Staleness — source-stale / context-stale / child-stale / degraded, refresh select, one successor per artifact
5. Agent access — semantic_memory and semantic_search include_memory, the 24 KB budget, the shedding order, and the two explicit empties
6. The 2026-08-17 Next.js scout — 704 billed calls by family, the 602-artifact coverage finding behind G18, and the Twenty memory gate numbers with their registered thresholds
7. Where to start reading — a treemap of the 24,587-line plane with a suggested reading order

Numbers are from the repository itself: wc -l on 2026-08-21, pipeline and lifecycle from docs/codebase-analysis-2026-08-17/08-scouting.md and the module headers, campaign figures from eval/results/next-root-params-types-2026-08-17.md and twenty-memory-gate-2026-08-09.md. Single self-contained HTML file, light and dark themes, no external dependencies beyond Google Fonts. Filed under presentations/ as a dated, non-normative artifact per that directory's README.