docs(ogar): calcify Northstar plan + fix(ogar-class-view): codex P2 stable order

[AdaWorldAPI]

What

Two follow-ups landing on top of the just-merged #77:

1. fix(ogar-class-view) — codex P2 stable order

known_class_ids() was borrowing HashMap's randomized iteration. Downstream bulk emitters using it for generated artifacts / drift snapshots would reorder across process runs without any schema change. Now iterates ogar_vocab::class_ids::ALL in codebook order; a regression test pins both stability and a second-call-equals-first invariant.

Replies the codex review on #77.

2. docs — calcify the Northstar plan

Rewrites docs/integration/CLASSVIEW-MATERIALIZATION-PLAN.md (125 → 486 LOC) to capture every architectural insight from this sprint so a future session does not drift. The plan is now the calcified single source of truth for this lane.

Section	What
§0	Picture — two-pipeline diagram, one codebook
§1	Six invariants — codebook calcified / N3 / Class-types vs ClassView-labels / ClassId == canonical_concept_id / ≤64 / mass-mail templates
§2	The two pipelines in detail (build-time codegen via &Class, run-time projection via ClassView), the muscle-memory loop, the UnifiedBridge layer above them
§3	The +5+5 follow-on PRs — T1–T5 askama templates + C1–C5 consumer wirings, each with three calibration gates
§4	Seven anti-patterns — drifts already explored and rejected, with the "why wrong" for each (the bicycle, bifurcation, template blowup, A2UI-in-now, codebook-skip, HashMap, unescaped keywords)
§5	Seven epiphanies captured verbatim (Rails words die / JSON PostgreSQL grind / class carries types / mass-mail templates / SoA knows its classview / dynamic doesn't go static at compile time / askama IS jinja-rs)
§6	Prior art lineage — WoA-rs (live antecedent), A2UI (roadmap T6), DUSK_Solution + MUIBridge (lineage only)
§7	Roadmap T6–T10 (A2UI, calibration aggregator, DOLCE wiring, 0x03XX+ promotions, commerce-arm convergence)
§8	Open PR map + status
§9	Pointer to convergence artifacts (redmine-canon / op-canon snapshots)

Sections 1, 4, 5 are mandatory reading before any future work in this lane.

Why this matters

The sprint surfaced architectural calls that took several round-trips to settle (e.g. the bicycle anti-pattern around typed_rows on ClassView; the bifurcation question; the binding-struct-is-just-a-bag insight). Calcifying them now stops a future session from re-deriving — or worse, drifting onto a rejected path.

Tests

cargo test -p ogar-class-view 9/9; workspace check + workspace test green.

Pair

PR #78 (ogar-render-askama) was force-rebased onto the new main (carries the codex P1 keyword-escape fix + this PR's Cargo.toml workspace member entry conflict-resolved cleanly).