Support MuLan, a plug-and-play language adapter to adapt existing diffusion model for up to 110+ languages without additional training

### Model/Pipeline/Scheduler description

```diff
# pip install mulankit
from diffusers import StableDiffusionPipeline
+ import mulankit

pipe = StableDiffusionPipeline.from_pretrained('Lykon/dreamshaper-8')
+ pipe = mulankit.transform(pipe, 'mulanai/mulan-lang-adapter::sd15_aesthetic.pth')
image = pipe('一只蓝色的🐶 in the 바다').images[0]
```

|一只蓝色的 🐶 in the 바다 (Dreamshaper-8)| レゴシュワルツェネッガー (SDXL-lightning)| 一只可爱的猫头鹰 (MVDream) | 海浪风景 (AnimateDiff) |
|--- | ---| --- | --- |
|  |  |  | |

MuLan supports

- Base models: Stable Diffusion 1.5, 2.1, XL, Pixart-Alpha/Sigma.
- Downstream models: ControlNet, LCM, LoRA, finetuned models and etc.
- Video models: AnimateDiff.
- 3D models: MVDream.

### Open source status

- [X] The model implementation is available.
- [X] The model weights are available (Only relevant if addition is not a scheduler).

### Provide useful links for the implementation

https://github.com/mulanai/MuLan
https://huggingface.co/mulanai/mulan-lang-adapter

Start with the StableDiffusionPipeline example and review the MuLan implementation at https://github.com/mulanai/MuLan and its weights at Hugging Face. Determine how support should cover the listed base, downstream, video, and 3D models. Done means MuLan can be integrated through diffusers for the described multilingual generation cases.

索引モデルが issue の本文から書いたものです。