[single file] Detect CLIP max_position_embeddings

[hlky]

What does this PR do?

In single file loading the CLIP config can be incorrect for checkpoints using something like LongCLIP due to the increased max_position_embeddings.

This PR detects the CLIP checkpoint's true max_position_embeddings and overrides the config then reinitializes the model if the detected value does not match.

Reproduction

huggingface-cli download opendiffusionai/xllsd-alpha0  --local-dir G:/xllsd-alpha0

python scripts/convert_diffusers_to_original_stable_diffusion.py --use_safetensors --model_path G:/xllsd-alpha0 --checkpoint_path G:/xllsd-alpha0.safetensors

from diffusers import StableDiffusionPipeline

pipe = StableDiffusionPipeline.from_single_file("G:/xllsd-alpha0.safetensors")

Fixes #11285

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[DN6]

Thanks 👍🏽