docs(osm-plan): route D-OSM-2 osmpbf consumer to lance-graph module, not ndarray stub

Addresses codex P2 on merged #473 (.claude/plans/cesium-osm-substrate-v1.md:91). The plan wired osmpbf v0.4 into the ndarray crates/cesium/src/osm_pbf.rs file, which D-OSM-1 explicitly declares dependency-free -- wrong repo/crate boundary, and it left the lance-graph Arrow/Lance emitter underspecified.

D-OSM-2 now directs the real osmpbf consumer + Arrow/Lance emitter to a lance-graph-side module (crates/lance-graph/src/ingest/osm_pbf.rs, new), reusing the D-OSM-1 carrier shapes + XYZ->TMS Y-flip helper; the ndarray cesium file stays the dependency-free D-OSM-1 stub. STATUS_BOARD D-OSM-2 row already attributed the deliverable to lance-graph (unchanged). AGENT_LOG: prepend this review-fix run.

https://claude.ai/code/session_013rjF2Dvo1DnBACpbpYSffE

Author: claude

.claude/board/AGENT_LOG.md

@@ -1,3 +1,15 @@
+## [Opus 4.8, main thread] cesium-osm-substrate-v1 review fix — D-OSM-2 crate boundary (codex P2 on merged #473)
+
+**Branch:** `claude/osm-pbf-consumer-boundary-fix` (off `main`). **New follow-up PR** (merged #473 review-fix, surfaced for visibility per user request; the other session owns the OGAR-side fixes).
+
+**Fix:** `.claude/plans/cesium-osm-substrate-v1.md` D-OSM-2 (line 91). Codex P2 (`chatgpt-codex-connector`, `#473#discussion_r3362274315`) flagged the plan wiring `osmpbf` v0.4 into the **ndarray** `crates/cesium/src/osm_pbf.rs` file — which D-OSM-1 explicitly declares dependency-free — putting the ingest dep in the wrong repo/crate boundary and leaving the lance-graph Arrow/Lance emitter underspecified. Re-pointed the real `osmpbf` consumer + Arrow/Lance emitter to a **lance-graph-side** module (`crates/lance-graph/src/ingest/osm_pbf.rs`, new), reusing the D-OSM-1 carrier shapes + XYZ→TMS Y-flip helper; the ndarray cesium file stays the dep-free D-OSM-1 stub. STATUS_BOARD D-OSM-2 row already attributed the deliverable to `lance-graph` (no change needed).
+
+**Scope:** lance-graph only. `ndarray#214` (D-OSM-1 stub) had no actionable review comments (CodeRabbit was rate-limited; no review threads). `OGAR#38/#39` are outside this session's repo scope → other session.
+
+**Tests:** none (docs/board only; `cargo` not invoked).
+
+---
+
 ## [Opus 4.7 / 1M ctx, main thread] cesium-osm-substrate-v1 — OSM as 6th Cesium ingest source class (cross-session coordination with OGAR)
 
 **Branch:** `claude/cesium-osm-substrate-v1` (new branch off `main`). **Files (this commit, lance-graph side):**

.claude/plans/cesium-osm-substrate-v1.md

@@ -88,7 +88,7 @@ Mirror the shape of `crates/cesium/src/arcgis_pbf.rs` (428 LOC, already shipped)
 
 ### D-OSM-2 — `osmpbf` consumer + Arrow RecordBatch emitter (lance-graph)
 
-Wire `osmpbf` v0.4 (b-r-u) into `crates/cesium/src/osm_pbf.rs`. Lazy-decoded parallelized PBF reader → per-element-type Arrow RecordBatch:
+Wire `osmpbf` v0.4 (b-r-u) into a **lance-graph-side** module — e.g. `crates/lance-graph/src/ingest/osm_pbf.rs` (new) — **not** the ndarray `crates/cesium/src/osm_pbf.rs` file, which stays the dependency-free D-OSM-1 stub. The `osmpbf` dependency and the Arrow/Lance emitter live entirely on the lance-graph side; this module consumes the D-OSM-1 carrier shapes (`OsmNode` / `OsmWay` / `OsmRelation` / `OsmTagList`) and the XYZ→TMS Y-flip helper, mapping each lazily-decoded, parallelized PBF element into a per-element-type Arrow RecordBatch:
 
 - `osm_nodes` table — `(id u64, lat f64, lon f64, qk_tms_path FixedSizeBinary(24), tags List<Struct<key,value>>)`
 - `osm_ways` table — `(id u64, ref_node_ids List<u64>, qk_tms_path FixedSizeBinary(24), tags List<Struct<key,value>>)`