test(acceptance): give the B256 case a copy folder of its own

Every case of a run shares one root, and the pdfs copy case before it
had already made "copies" there, so the B256 case failed on its mkdir
with EEXIST before it checked anything.

crates/pdfs-fuse/src/lib.rs

@@ -7312,15 +7312,13 @@ mod size_upgrade_tests {
     fn a_listing_waits_for_the_whole_size_batch_but_not_forever() {
         let slot = Arc::new(SizeUpgrade::default());
         let worker = slot.clone();
-        // Read before the spawn: the worker's sleeps can start before this
-        // thread runs again, which made the bound below fail on a busy runner.
-        let started = Instant::now();
         let t = std::thread::spawn(move || {
             std::thread::sleep(Duration::from_millis(30));
             worker.chunk_done();
             std::thread::sleep(Duration::from_millis(60));
             worker.finish();
         });
+        let started = Instant::now();
         slot.wait_finished(started + Duration::from_secs(5));
         // Past the first chunk, up to the end of the batch.
         assert!(started.elapsed() >= Duration::from_millis(90));

scripts/fuse-acceptance.py

@@ -787,7 +787,9 @@ def test_server_copy_never_copies_an_old_version(ctx: Context) -> None:
     root = ctx.root
     path = root / "being-written.txt"
     write_durable(path, b"first version")
-    dest = root / "copies"
+    # Its own folder: every case of a run shares the root, and "pdfs copy"
+    # already made "copies" there.
+    dest = root / "b256-copies"
     dest.mkdir()
     daemon.wait_for_queue()