Use `tuple` instead of `list` in `gen_kwargs` to avoid `IterableDataset.shuffle` error

Same change as mariosasko's PR here: https://huggingface.co/datasets/fashion_mnist/discussions/3

mnist.py

@@ -87,14 +87,14 @@ class MNIST(datasets.GeneratorBasedBuilder):
             datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={
-                    "filepath": [downloaded_files["train_images"], downloaded_files["train_labels"]],
+                    "filepath": (downloaded_files["train_images"], downloaded_files["train_labels"]),
                     "split": "train",
                 },
             ),
             datasets.SplitGenerator(
                 name=datasets.Split.TEST,
                 gen_kwargs={
-                    "filepath": [downloaded_files["test_images"], downloaded_files["test_labels"]],
+                    "filepath": (downloaded_files["test_images"], downloaded_files["test_labels"]),
                     "split": "test",
                 },
             ),