Standardize examples under more TensorFlow idioms

[dustinvtran]

This PR standardizes Edward examples to follow seven TensorFlow idioms:

1. main() stores script code followed by tf.app.run();
2. tf.flags for global configurations;
3. tf.get_variable for variable definitions;
4. tf.layers for neural net layers;
5. net as the input/output variable name in layer compositions;
6. Use (...) for multi-line variable assignments rather than \ ;
7. In-place operators such as /= when possible.

Note. Some cases can be hard to find a consistent idiom. For example, official TF examples vary among absl, tf.flags, tf.app.flags, argparse; flags are added at both top of script and bottom within if __name__ == "__main__": clause; single quote vs double quotes for phrases; main(unused_argv) vs main(_).

I used my own judgments and worked off examples such as:

• https://github.com/tensorflow/models/blob/master/tutorials/rnn/ptb/ptb_word_lm.py
• https://github.com/tensorflow/models/blob/master/samples/core/get_started/premade_estimator.py
• https://github.com/tensorflow/tpu-demos/blob/master/cloud_tpu/models/resnet_garden/resnet_main.py
• https://github.com/tensorflow/tensorflow/blob/master/tensorflow/contrib/eager/python/examples/gan/mnist.py

Note. I did not do anything to the code except perform the conversions. A future PR can follow more TensorFlow idioms. For example: rename flags to be self-documenting over Greek (e.g., batch_size overM); use broadcasting over explicit kwargs in random variables; sample_shape.

todos

• update main pages
• update examples/
• update notebooks/ and docs/tex/
• update tests/