fix: use dynamic label names for inference to match training dataset

Previously, the inference section used a hardcoded list of labels:
['no', 'yes', 'down', 'go', 'left', 'up', 'right', 'stop'],
which may not match the order of class names in `train_ds.class_names`.
This could lead to incorrect labeling in prediction visualizations.

This commit replaces the hardcoded labels with `label_names` from the training
dataset to ensure the model's outputs are correctly mapped to their respective labels.

Author: Mani-Babapour

site/en/tutorials/audio/simple_audio.ipynb

@@ -825,7 +825,7 @@
         "x = x[tf.newaxis,...]\n",
         "\n",
         "prediction = model(x)\n",
-        "x_labels = ['no', 'yes', 'down', 'go', 'left', 'up', 'right', 'stop']\n",
+        "x_labels = label_names\n",
         "plt.bar(x_labels, tf.nn.softmax(prediction[0]))\n",
         "plt.title('No')\n",
         "plt.show()\n",