Object Detection support

• bug fixes
• refactores
• moved modules to tensorpipe
• added support for Bounding boxes/Object Detection data.

Categorical Data with optimizations and numpy function injection.

Custom numpy injection in the data pipeline with optimization
SOTA augmentations with optimizations.
Stable pipeline with error handling.
currently only categorical data is being supported.

Classification support

Features supported and tested:

1. All tf.image and tfa.image augmentations are supported.
2. GridMask is supported and tested.
3. Tested entire pipeline on categorical dataset.
4. Custom functions are not yet supported.
5. Augmentation pipeline is configurable through config file.