perf: optimize indexing speed with batching and incremental sync

- Batching BDD: group INSERT/UPDATE operations in batches of 100 using UNNEST
- Incremental MeiliSearch: only sync books modified since last sync
- Optimized fingerprint: use only size+mtime+filename (100x faster)
- Increased DB connections from 5 to 20
- Reduced progress update frequency (every 1s or 10 files)
- Add sync_metadata table to track last MeiliSearch sync

infra/migrations/0007_add_sync_metadata.sql

@@ -0,0 +1,6 @@
+CREATE TABLE IF NOT EXISTS sync_metadata (
+    id INTEGER PRIMARY KEY,
+    last_meili_sync TIMESTAMPTZ
+);
+
+INSERT INTO sync_metadata (id, last_meili_sync) VALUES (1, NULL) ON CONFLICT DO NOTHING;