[feature] Tiny modification to enable OFT for finetuning embedding layers

[zqiu24]

Hi there,

I would like to merge this feat: Add OFT support for Embedding layers

• Implement OFT Embedding layer with output rotation
• Add merge/unmerge support for embedding weights
• Override _available_adapters and reset_oft_parameters for embedding-specific handling
• Tested with training script, matches LoRA performance"

Hope for a smooth integration. Thank you so much. :D

Best,