Move TF1 hashing example out of code cell.

A planned change in TensorFlow will disable hashing
of tensors in eager execution mode even when
numpy-style equality is disabled
(the TensorFlow 1 default).

This will break the moved example. Since we cannot
disable and re-enable eager execution after the
beginning of the program, this example is no longer
runnable under TF2 - even with the compatibility
layer.

As noted in the edited documentation, `set` requires
its elements to be hashable - since this example runs
under TF2 and we will no longer allow hashing of tensors
in eager execution mode full-stop, the code _is_ valid
TF1 code but cannot be run with the proposed changes to TF2.

PiperOrigin-RevId: 520100823

Author: Spyboticsguy

site/en/guide/migrate/tf1_vs_tf2.ipynb

@@ -1027,28 +1027,12 @@
       },
       "source": [
         "### Hashing tensors and variables\n",
-        "With TF1.x behaviors you used to be able to directly add variables and tensors to data structures that require hashing, such as `set` and `dict` keys."
-      ]
-    },
-    {
-      "cell_type": "code",
-      "execution_count": null,
-      "metadata": {
-        "id": "fxi7EgKdBpd5"
-      },
-      "outputs": [],
-      "source": [
-        "tf.compat.v1.disable_tensor_equality()\n",
+        "With TF1.x behaviors you used to be able to directly add variables and tensors to data structures that require hashing, such as `set` and `dict` keys.\n",
+        "```python\n",
         "x = tf.Variable(0.0)\n",
-        "set([x, tf.constant(2.0)])"
-      ]
-    },
-    {
-      "cell_type": "markdown",
-      "metadata": {
-        "id": "LFgjR-PSajVQ"
-      },
-      "source": [
+        "set([x, tf.constant(2.0)])\n",
+        "```\n",
+        "\n",
         "However, in TF2 with tensor equality enabled, tensors and variables are made unhashable due to the `==` and `!=` operator semantics changing to value equality checks."
       ]
     },