Add random_shear processing layer

[shashaka]

I have implemented the transform_images method in the RandomShear image processing layer. If this pull request is accepted, I can proceed with implementing the transform_bounding_boxes method as well.

here is my gist

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 78.88889% with 19 lines in your changes missing coverage. Please review.

Project coverage is 81.94%. Comparing base (6ce93a4) to head (010a402).

Files with missing lines	Patch %	Lines
.../preprocessing/image_preprocessing/random_shear.py	78.16%	10 Missing and 9 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master   #20702      +/-   ##
==========================================
- Coverage   81.95%   81.94%   -0.01%     
==========================================
  Files         547      548       +1     
  Lines       51036    51126      +90     
  Branches     7893     7905      +12     
==========================================
+ Hits        41826    41897      +71     
- Misses       7285     7295      +10     
- Partials     1925     1934       +9     

Flag	Coverage Δ
keras	81.77% <78.88%> (-0.01%)	⬇️
keras-jax	63.99% <78.88%> (+0.02%)	⬆️
keras-numpy	58.93% <74.44%> (+0.02%)	⬆️
keras-openvino	29.89% <27.77%> (-0.01%)	⬇️
keras-tensorflow	64.68% <78.88%> (+0.02%)	⬆️
keras-torch	64.06% <78.88%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[fchollet]

LGTM -- thank you! 👍