perf(classifier): raise scan concurrency 16 -> 32

The scan is I/O-bound on cloud-sync / network mounts (OneDrive, Samba) where
each directory read is a high-latency round-trip. More in-flight reads hide
that latency on the many-folders case. (A single large folder is still
enumerated one entry at a time by the File System Access API and can't be
parallelized.)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

classifier/js/scanner.js

@@ -16,11 +16,15 @@
     let scanGen = 0;                 // bumped per scan; stale workers bail
     let scanStats = null;            // { folders, files, current, done, startedAt }
     let renderTimer = null;          // throttle for progressive re-render
-    // How many directory reads to keep in flight at once. The scan is
-    // I/O-bound (each readdir is a network round-trip to the share), so the
-    // lever is parallel in-flight reads, not CPU threads — raise this if the
-    // share tolerates more concurrency.
-    var SCAN_CONCURRENCY = 16;
+    // How many directories to read in flight at once. The scan is I/O-bound —
+    // each readdir is a round-trip to the backing store (cloud-sync / network
+    // mounts like OneDrive or Samba have high per-op latency), so the lever is
+    // parallel in-flight reads, not CPU threads. This only helps the
+    // many-folders case; a single fat folder is enumerated one entry at a time
+    // by the File System Access API and can't be parallelized. Raise it if the
+    // store tolerates more concurrency; too high risks cloud-provider
+    // throttling.
+    var SCAN_CONCURRENCY = 32;
 
     function scheduleRender() {
         if (renderTimer) return;