Bacterial Deepfakes: Generating Synthetic Microscopy Data to Improve Adaptability of Deep Learning-Based Segmentation Models