update augmentations to pass and return bbox and label transformation…

[bernardev254]

Summary

This PR updates the augmentations() function in hackathon/objectdetection.py to support bounding box and label augmentation using create_global_crops.

Changes

• Passes bboxes, labels, and num_classes into create_global_crops.
• Receives and handles the transformed bounding boxes and labels (as crops_heatmaps).
• Stores the augmented object data in the returned dictionary under the "objects" key.
• Keeps local_crops logic optional and unchanged.

Motivation

This enables consistent transformation of object detection targets (bounding boxes and labels) during augmentation, aligning with the changes made to create_global_crops and preparing the pipeline for training with augmented samples.
closes #3