fix(dogfood): document-centric fetch_span + remove get_asset_by_workspace_path

assets.workspace_path is INTENTIONALLY 'last-registered path' for twin
files (identical content at different paths share one asset row PK'd by
blake3 content hash). PR #146 made try_skip_unchanged document-centric;
PR #149 made reset --orphans-only document-centric; this PR removes the
last caller of get_asset_by_workspace_path (fetch.rs:193 in fetch_span,
which used it to reject PDF/audio media — for twins this could read the
wrong asset's media_type and pick the wrong branch).

Replaced with the natural 2-step lookup: get_document_by_workspace_path
(PR #146) → doc.source_asset_id → get_asset (NEW trait method, asset_id
is PRIMARY KEY so flip-flop-immune by construction).

Then removed get_asset_by_workspace_path trait method + SqliteStore impl
— 0 callers after the refactor.

UPSERT doc-comment refreshed in store.rs to make the 'last-registered'
semantics explicit so future readers don't try to 'fix' the flip-flop.

Dogfood follow-up (PR #142 1B + multi-root corpus).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

crates/kebab-store-sqlite/src/documents.rs

@@ -264,6 +264,28 @@ impl kebab_core::DocumentStore for SqliteStore {
         }))
     }
 
+    fn get_asset(
+        &self,
+        id: &kebab_core::AssetId,
+    ) -> Result<Option<kebab_core::RawAsset>> {
+        let conn = self.lock_conn();
+        let result = conn.query_row(
+            r#"SELECT
+                asset_id, source_uri, workspace_path, media_type,
+                byte_len, checksum, storage_kind, storage_path,
+                discovered_at
+            FROM assets
+            WHERE asset_id = ?"#,
+            rusqlite::params![id.0.as_str()],
+            asset_from_row,
+        );
+        match result {
+            Ok(asset) => Ok(Some(asset)),
+            Err(rusqlite::Error::QueryReturnedNoRows) => Ok(None),
+            Err(e) => Err(e.into()),
+        }
+    }
+
     fn get_asset_by_workspace_path(
         &self,
         path: &kebab_core::WorkspacePath,
@@ -632,7 +654,8 @@ fn rows_optional<T>(err: rusqlite::Error) -> rusqlite::Result<Option<T>> {
 
 /// Reconstruct a [`kebab_core::RawAsset`] from one `assets` row.
 /// Row mapper for `RawAsset`. Column names are self-documenting; the
-/// SELECT in [`DocumentStore::get_asset_by_workspace_path`] must include
+/// SELECTs in [`DocumentStore::get_asset`] and
+/// [`DocumentStore::get_asset_by_workspace_path`] must both include
 /// all nine columns by their schema names.
 fn asset_from_row(row: &rusqlite::Row<'_>) -> rusqlite::Result<kebab_core::RawAsset> {
     use std::path::PathBuf;

crates/kebab-store-sqlite/src/store.rs

@@ -652,6 +652,20 @@ pub fn purge_deleted_workspace_path(
 /// path (which has bytes + computed `storage_kind/path`) and the
 /// `DocumentStore::put_asset` path (which only has the `RawAsset` and
 /// reads `storage_kind/path` from `asset.stored`).
+///
+/// **`assets.workspace_path` is "last-registered path" semantics for
+/// twin files** (two source files with identical content share one
+/// `assets` row keyed on `asset_id = blake3(content)`). Each ingest
+/// of either twin overwrites `workspace_path` with whichever path was
+/// seen most recently — this is intentional and correct after PR #146
+/// made `try_skip_unchanged` document-centric (uses
+/// `get_document_by_workspace_path`, not `get_asset_by_workspace_path`)
+/// and PR #149 made `reset --orphans-only` document-centric too.
+/// Do NOT "fix" the flip-flop by adding a UNIQUE constraint on
+/// `workspace_path` in the `assets` table — twin de-dup is load-bearing.
+/// When you need media_type for a known document, use the 2-step lookup
+/// `get_document_by_workspace_path` → `doc.source_asset_id` →
+/// `get_asset(asset_id)` so the result is twin-safe.
 pub(crate) fn upsert_asset_row(
     conn: &Connection,
     asset: &kebab_core::RawAsset,