Fixed few code lines in this API example

Updated `axis` to `axes` as per  tf.nn.moments() actual syntax and added full alias name (tfp.mcmc.) to effective_sample_size() to execute the example successfully. I have replicated and fixed the code error of this API example in this [gist](https://colab.sandbox.google.com/gist/RenuPatelGoogle/9a528802d6e52e46ad9713c6391bbc5c/tfp-mcmc-effective_sample_size.ipynb#scrollTo=pW_kQ2Px3VLs) for your reference.

Author: RenuPatelGoogle

tensorflow_probability/python/mcmc/diagnostic.py

@@ -160,13 +160,14 @@ def effective_sample_size(states,
         target_log_prob_fn=target.log_prob,
         step_size=0.05,
         num_leapfrog_steps=20))
-  states.shape
+  print(states.shape)
   ==> (1000, 2)
 
   ess = effective_sample_size(states, filter_beyond_positive_pairs=True)
-  ==> Shape (2,) Tensor
+  print(ess.shape)
+  ==> (2,)
 
-  mean, variance = tf.nn.moments(states, axis=0)
+  mean, variance = tf.nn.moments(states, axes=0)
   standard_error = tf.sqrt(variance / ess)
   ```