Adding improved ResNets to `tf.keras.applications`

I think we can agree that the importance of ResNets in the Deep Learning community is paramount. Since their inception, they have gone through several improvements most of them leading to increased performance on the ImageNet-1k dataset.

This thread seems contributions from the community for adding ResNetRS models to tf.keras.applications.

This has been merged in a9b4cb6 and tensorflow/tensorflow@e9c5ec2.

Please refer to the following ResNet-RS models for Keras.

ResNetRS50(...): Instantiates the ResNetRS50 architecture.

ResNetRS101(...): Instantiates the ResNetRS101 architecture.

ResNetRS152(...): Instantiates the ResNetRS152 architecture.

ResNetRS200(...): Instantiates the ResNetRS200 architecture.

ResNetRS270(...): Instantiates the ResNetRS270 architecture.

ResNetRS350(...): Instantiates the ResNetRS350 architecture.

ResNetRS420(...): Instantiates the ResNetRS420 architecture.

Thank you.