[WIP] Attempt at cleaning backbone utils

[zucchini-nlp]

What does this PR do?

Backbone utils logic looks a bit complex and we have some args that can be easily merged into one (backbone_config and bacbone_kwargs)

I will try to remove stuff, specifically:

• Disallow backbone_kwargs and move it into backbone_config in each model's config (e.g. backbone_config = CONFIG_MAPPINGmodel_type)
• Remove use_pretrained_backbone and infer it from backbone_checkpoint is not None. I think loading a pretrained or random weights should be dependent on whether you call from_pretrained(model_id) or __init__(config)
• See if config.use_timm_backbone is necessary or if calling AutoBackbone.from_pretrained() can automatically infer model type from config

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: auto, conditional_detr, d_fine, depth_anything, grounding_dino, mask2former, omdet_turbo, oneformer, table_transformer, tvp, upernet, vitmatte

[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.

[github-actions]

View the CircleCI Test Summary for this PR:

https://huggingface.co/spaces/transformers-community/circle-ci-viz?pr=43323&sha=71f301