docs(documents): TODO for SQL aggregation at 50k rows (#385)

Documents the in-memory aggregation trade-off in getDensity so the next
perf audit knows the row-count threshold at which to revisit. Addresses
Markus's review concern.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

backend/src/main/java/org/raddatz/familienarchiv/document/DocumentService.java

@@ -139,6 +139,11 @@ public class DocumentService {
      * with {@code findAll(spec)} and the archive size (≈5k docs) keeps this
      * well under the 200ms p95 target. Cache-Control: max-age=300 on the
      * controller layer absorbs repeated browse loads.
+     *
+     * <p>TODO: revisit when {@code documents > 50k} — at that scale move the
+     * aggregation into SQL (GROUP BY TO_CHAR(meta_date, 'YYYY-MM')) and accept
+     * that the criteria/specification surface needs a parallel native-query
+     * path. Surface this in the next perf audit.
      */
     public DocumentDensityResult getDensity(
             String text, UUID sender, UUID receiver,