OneTrainer lora format for ideogram4 ✨

### Feature Idea

OneTrainer just added Ideogram4 support, bringing with it split q k v loras!

Would be great if they could be supported.

Example lora:

https://huggingface.co/bertbobson/temp_stuff/blob/main/Bad.safetensors

if I rename the keys to match the expected comfyui naming scheme, these are not loaded per block:

```
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_k.alpha
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_k.lora_A.weight
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_k.lora_B.weight
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_out.0.alpha
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_out.0.lora_A.weight
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_out.0.lora_B.weight
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_q.alpha
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_q.lora_A.weight
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_q.lora_B.weight
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_v.alpha
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_v.lora_A.weight
[WARNING] lora key not loaded: diffusion_model.layers.0.attention.to_v.lora_B.weight

```

### Existing Solutions

_No response_

### Other

I love diffusers