Stablepy

**Full Changelog**: https://github.com/R3gm/stablepy/compare/0.5.1...0.5.2

- After generation, PIL images are now returned along with sublists containing the seeds, image paths, and metadata with the parameters used in generation. `[pil_image], [seed, image_path, metadata]`
- The use of `image_previews=True` has been improved, and now preview images can be obtained during the generation steps using a generator. For more details, refer to the Colab notebook.

- Fix LoRA SDXL compatibility.
- Latent upscaler and variants.
- Perturbed Attention Guidance (PAG) enhances image generation quality without the need for training.
- Multiple images for one FaceID adapter.
- ControlNet for SDXL: MLSD, Segmentation, Normalbae.
- ControlNet "lineart_anime" task accessible and able to load a model different from the "lineart" task.
- ControlNet Tile and Recolor for SD1.5 and SDXL ("tile" replaces the previous task called "sdxl_tile_realistic").

- Fix the aspect ratio for SDXL inpaint

- IP Adapter with the variants FaceID and Instant-Style
- New samplers
- Appropriate support for SDXL safetensors models
- ControlNet for SDXL: OpenPose, Canny, Scribble, SoftEdge, Depth, LineArt, and SDXL_Tile_Realistic
- New variant prompt weight with emphasis
- ControlNet pattern for SD1.5 and SDXL
- ControlNet Canny now needs the preprocessor_name="Canny"
- Similarly, ControlNet MLSD requires the preprocessor_name="MLSD"
- Task names like "sdxl_canny" have been changed to "sdxl_canny_t2i" to refer to the T2I adapter that uses them.