[Feature Request] Add LoRA loading support for HunyuanVideo 1.5

Dieses Issue hat noch niemand übernommen.

**What does this feature request propose?**

HunyuanVideo 1.5 was recently added to diffusers with two pipelines:
- `HunyuanVideo15Pipeline` (text-to-video)
- `HunyuanVideo15ImageToVideoPipeline` (image-to-video)

However, neither pipeline has LoRA support. Both classes currently inherit only from `DiffusionPipeline` with no `LoraLoaderMixin`.

HunyuanVideo v1 already has `HunyuanVideoLoraLoaderMixin` covering its pipelines, but that mixin explicitly targets `HunyuanVideoTransformer3DModel`. HunyuanVideo 1.5 uses a distinct `HunyuanVideo15Transformer3DModel`, so it cannot reuse the v1 mixin directly.

**Proposed implementation**

- Add `HunyuanVideo15LoraLoaderMixin` to `src/diffusers/loaders/lora_pipeline.py`, modelled after the existing `HunyuanVideoLoraLoaderMixin` but targeting `HunyuanVideo15Transformer3DModel`
- Wire both `HunyuanVideo15Pipeline` and `HunyuanVideo15ImageToVideoPipeline` to inherit from it
- Export the new mixin from `src/diffusers/loaders/__init__.py`
- Add tests mirroring the existing HunyuanVideo LoRA test coverage

**Why is this needed?**

HunyuanVideo has a large and active fine-tuning community. LoRA training for 1.5 is already happening in tools like ai-toolkit and ComfyUI. Without a standard loader in diffusers, users have no clean path to load those LoRAs through the library.

**Who can review?**
`@sayakpaul` `@DN6` `@yiyixuxu`

**Are you willing to contribute?**
Yes, happy to open a PR if the approach looks reasonable.

Beginne mit HunyunVideoLoraLoaderMixin in src/diffusers/loaders/lora_pipeline.py und vergleiche die Pipeline-Klassen von HunyuanVideo 1.5. Verfolge die vorhandenen HunyuanVideo-LoRA-Tests und füge anschließend entsprechende Abdeckung für beide Pipelines hinzu. Überprüfe außerdem, dass das neue Mixin aus src/diffusers/loaders/__init__.py exportiert wird; fertig ist die Arbeit, wenn beide Pipelines HunyuanVideo 1.5-LoRAs über den standardmäßigen Loader laden können.

Vom Indexierungsmodell aus dem Issue-Text verfasst.