[Example] Added WGAN (Wasserstein GAN) to keras-gan examples

[chibihime]

This PR adds a new WGAN implementation to the examples/keras-gan folder. The model uses the Wasserstein loss with a clipped critic and provides more stable training compared to standard GANs.

[chibihime]

This PR introduces a Keras implementation of a Wasserstein GAN (gan-wgan.py) under examples/keras-gan/. The model uses Wasserstein loss with a clipped critic for more stable training, as proposed in the original WGAN paper. This addition complements the existing GAN and DCGAN examples, and can be helpful for learners exploring alternative GAN architectures.

Let me know if any changes are needed!