Add training pipeline for AE CNN models

[gc9591]

• Implemented step 0a to compute spectrograms from raw HDF5 signals and save as multichannel TIFFs.
• Created step 1 to generate semantic labels using a trained model.
• Developed step 1a (alternative) to generate semantic labels from pre-computed spectrograms.
• Added step 2a to create COCO dataset from semantic labels.
• Implemented step 2b to convert COCO dataset to YOLO format.
• Established step 3 for training Faster R-CNN and Mask R-CNN models.
• Added training scripts for YOLOv8 detection and segmentation.
• Created step 4 for saving model checkpoints.
• Introduced necessary utility functions and configurations for dataset handling and model training.