*   Remove `tf.contrib.data.Iterator.from_dataset()` method. Use
    `Dataset.make_initializable_iterator()` instead.
*   Remove seldom used and unnecessary `tf.contrib.data.Iterator.dispose_op()`.
*   Reorder some TF-GAN loss functions in a non-backwards compatible way.

## Known Issues

*   In Python 3, `Dataset.from_generator()` does not support Unicode strings.

so those were implemented using // macros to compute the equivalent values. // 3. The LVX, STVX instructions on ppc64 require // 16 byte alignment of the data. To avoid that // requirement, data is loaded using LXVD2X and // STXVD2X with VPERM to reorder bytes correctly. // I have identified some areas where I believe // changes would be needed to make this work for big // endian; however additional changes beyond what I // have noted are most likely needed to make it work. // - The string used with...