Enhance MLOpsPython to support image dataset and Tensorflow/Keras models

[liupeirong]

MLOpsPython has comprehensive devops and mlops pipelines that support the full life cycle of machine learning projects. However, it only demonstrates a scikit-learn model with tabular data. Image data typically requires reshaping the images in a separate data processing pipeline and a different type of Azure ML dataset. Computer vision models based on Tensorflow/Keras also requires training and inferencing environments different from scikit-learn.

This item builds and deploys the flower image classification example in the Tensorflow tutorial. Specifically, it modifies MLOps in the following ways -

• Has a separate data processing pipeline to reshape the images.
• Supports images as Azure ML dataset.
• Trains and deploys a Tensorflow/Keras based image classification model.
• Preserves the devops and mlops pipelines in MLOpsPython.