Feat: Implement ImageDataGenerator for real-time data augmentation

[debug-soham]

Description:

This PR adds the ImageDataGenerator class in pydeepflow/preprocessing.py to enable real-time image data augmentation (rotation, shift, zoom, flip). This enhances the CNN capabilities by helping prevent overfitting and improving model robustness.

Key Changes:

• New File: pydeepflow/preprocessing.py containing the ImageDataGenerator.
• Modified: pydeepflow/model.py - Updated fit() methods in Multi_Layer_ANN and Multi_Layer_CNN to accept the generator. Fixed pooling layer __init__ for integer pool_size.
• Modified: pydeepflow/__init__.py - Exported ImageDataGenerator.

Test

The new feature can be tested by creating an instance of the ImageDataGenerator and passing its flow method to the model's fit function. All existing tests continue to pass, ensuring no regressions were introduced.

Closes #117

[ravin-d-27]

Amazing! Thank you!