Wrap specific padding behavior for convolution layer in tf.keras

[Zhaopudark]

Description

more details in #2674

Extend padding behavior of a keras convolution layer.
For example, a user can get a convolution layer with "same" padding in shape-wise and "reflect" padding in numerical-wise simultaneously. Such as:

x = tf.constant([1.,2.,3.,4.,5.],shape=[1,5,1])
conv1d_ = tf.keras.layers.Conv1D(filters=1, kernel_size=[2,]*1, strides=(1,)*1, padding="same")
conv1d = SpecificConvPad(conv1d_, padding_mode='constant',padding_constant=1)
y = conv1d(x)

Type of change

• Additional Testing
• New Layer and the changes conform to the layer contribution guidelines

Checklist:

• I've properly formatted my code according to the guidelines
  • By running Black + Flake8
  • By running pre-commit hooks
• I have added tests that prove my fix is effective or that my feature works

How Has This Been Tested?

Shape-wise padding behavior has been tested to ensure to be consistent with original padding behavior.

[bhack]

Thanks for the proposal, can you try to open a ticket and propose this in Keras directly:
https://github.com/keras-team/keras

[bhack]

I think you need to contribute/propose this in Keras or Keras-cv repos