feat: index the text of attached PDFs

- jz_bm_material_file_texts caches the pdftotext output per file
- the River job jz.bm.extract_material_text extracts and resyncs
- the management multipart attach queues it inside its transaction
- file_texts in the document and a file snippet with its name

controllers/api/materials.go

@@ -157,15 +157,21 @@ func (h *Handlers) Materials(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 	if cands != nil {
-		attachSnippets(out, rows, cands.Hits, viaMCP(r))
+		files, err := snippetFiles(db, materialIDs(rows), cands.Hits)
+		if err != nil {
+			writeOpaque500(w)
+			return
+		}
+		attachSnippets(out, rows, cands.Hits, files, viaMCP(r))
 	}
 	writeJSONPage(w, lagoon.Paginate(out, page, perPage, total))
 }
 
-// attachSnippets sets the snippet of each list row from its Hit. rows are
+// attachSnippets sets the snippet of each list row from its Hit and, for a
+// file snippet, the row's indexed files. rows are
 // the rows SQL returned (they passed the re-gate), in the order of out; the
 // Hit of an id SQL did not return is never read.
-func attachSnippets(out []materialListJSON, rows []models.Material, hits map[uint]beachcomber.Hit, mcp bool) {
+func attachSnippets(out []materialListJSON, rows []models.Material, hits map[uint]beachcomber.Hit, files map[uint][]models.IndexedFile, mcp bool) {
 	for i := range out {
 		if i >= len(rows) || rows[i].ID != out[i].ID {
 			continue
@@ -174,10 +180,19 @@ func attachSnippets(out []materialListJSON, rows []models.Material, hits map[uin
 		if !ok {
 			continue
 		}
-		out[i].Snippet = snippetValue(pickSnippet(hit, rows[i]), mcp)
+		out[i].Snippet = snippetValue(pickSnippet(hit, rows[i], files[rows[i].ID]), mcp)
 	}
 }
 
+// materialIDs are the ids of rows, in order.
+func materialIDs(rows []models.Material) []uint {
+	ids := make([]uint, 0, len(rows))
+	for _, m := range rows {
+		ids = append(ids, m.ID)
+	}
+	return ids
+}
+
 // Material answers GET /materials/{id}: one published material the caller
 // may see, with its body, tags and files. A draft, a future publication or
 // another user's private material is a 404. can_manage tells the author (a