Support for T5 embeddings

Currently, textual inversion embeddings for multiple text encoders only works with SDXL - https://github.com/lllyasviel/stable-diffusion-webui-forge/blob/main/modules/textual_inversion/textual_inversion.py#L291

I'd like to be able to use an embedding for T5 as well. ComfyUI looks for the key in the safetensor dict 't5xxl'. I'm able to use embeddings with Flux in Comfy that have the keys ['clip_l', 't5xxl'].

Start in modules/textual_inversion/textual_inversion.py around line 291, where multi-text-encoder textual inversion handling is limited to SDXL. Trace how safetensor keys are selected and compare the existing behavior with Flux embeddings containing 'clip_l' and 't5xxl'. Done means embeddings with the 't5xxl' key can be used with T5-based models.

Written by the indexing model from the issue text.