Import JAX substrate through tfp.substrates.jax instead of deprecated

experimental.substrates.
Also, have the colab binary depend on both the TensorFlow and JAX substrates
since various notebooks use each.

PiperOrigin-RevId: 613768759

tensorflow_probability/examples/jupyter_notebooks/TFP_Release_Notebook_0_11_0.ipynb

@@ -147,7 +147,7 @@
         "config.update('jax_enable_x64', True)\n",
         "\n",
         "def demo_jax():\n",
-        "  from tensorflow_probability.python.experimental.substrates import jax as tfp\n",
+        "  import tensorflow_probability.substrates.jax as tfp\n",
         "  tfd = tfp.distributions\n",
         "  tfb = tfp.bijectors\n",
         "\n",
@@ -2123,4 +2123,4 @@
       "outputs": []
     }
   ]
-}
+}