Batch norm update bug

https://github.com/carpedm20/DCGAN-tensorflow/blob/60aa97b6db5d3cc1b62aec8948fd4c78af2059cd/model.py#L223

In here that you're intending to just update the discriminator, but you are updating the batch_norm parameters of the generator too.

same holds for here:
https://github.com/carpedm20/DCGAN-tensorflow/blob/60aa97b6db5d3cc1b62aec8948fd4c78af2059cd/model.py#L232

You are updating self.g_loss where inside it uses the discriminator weights that are instantiated with `train=True`.

Start in model.py at the referenced lines 223 and 232, and trace how train=True affects the generator and discriminator batch-normalization parameters during each loss update. Verify the intended discriminator-only update behavior and confirm that the generator loss does not update the wrong batch-normalization state.

Written by the indexing model from the issue text.