[GSoC] Autoencoder migration

[Vishal-V]

• Migrated the autoencoder_runner.py and Autoencoder.py to Tensorflow 2.0
• The runner uses tf.data.Dataset pipeline to feed the batches and tf.keras.datasets to load the initial data and all necessary migrations in runner have been tested
• The backpropagation is in Eager mode with tf.GradientTape
• Wrote the Encoder and Decoder as subclasses of tf.keras.layers.Layer and the Autoencoder as a subclass of tf.keras.Model
• Tested the model with both Eager mode training as well as Graph mode training with model.fit.

[MarkDaoust]

This looks fine, but we might not be able to keep the stand alone copy of the code in tensorflow_examples/models, I'll check with the owners on the merge.

[lamberta]

Thanks for the contribution but this community/ directory has been deprecated. We're no longer accepting pull requests here. Not sure where the rest of this PR stands, though.