ALM base model class

[eustlb]

What does this PR do?

Fix a discrepancy in ALMs design compared to VLMs that for most of them don't have a base model class because the use the causal model for the text model directly, while Llava style (which is indeed more aligned with the lib philosophy) is rather:
XxModel: encoder + projector + language base model backbone
XxForConditionalGeneration: XxModel + lm_head

This is motivated to simplify changes required for vLLM compatibility as seen in #39330

how to ensure this is BC?

• weight loading: thanks to the dynamic weight loader and conversion mapping
• TODO further verifs

[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.