Add support for JoyAI-Video-Edit

### Model/Pipeline/Scheduler description

[JoyAI-Video-Edit](https://github.com/jd-opensource/JoyAI-Video-Edit) is an instruction-guided video editing model based on the JoyAI streaming architecture.

It performs causal, chunk-wise video denoising with a per-layer KV cache. Each chunk attends to previously denoised chunks and may additionally attend to an optional reference image.

Prompt embeddings are produced by the external [MiMo-VL](https://huggingface.co/XiaomiMiMo/MiMo-VL-7B-RL-2508) model.

### 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

- Reference implementation: https://github.com/jd-opensource/JoyAI-Video-Edit
- Original checkpoint: https://huggingface.co/jdopensource/JoyAI-Video-Edit
- MiMo-VL: https://huggingface.co/XiaomiMiMo/MiMo-VL-7B-RL-2508

No repository files, tests, or entry points are named. Start by comparing the linked reference implementation and checkpoint with existing video model integrations in diffusers, then determine how MiMo-VL prompt embeddings, causal KV caching, chunk-wise denoising, and optional reference images fit the project. Done means JoyAI-Video-Edit is supported using the linked model resources.

Vom Indexierungsmodell aus dem Issue-Text verfasst.