`eager_attention_forward` and `repeat_kv` code duplication

I see the two functions appear in a lot of places in the code base. Shall we unify them into a single place?

And can we treat `eager_attention_forward` as another option in [`ALL_ATTENTION_FUNCTIONS`](https://github.com/huggingface/transformers/blob/main/src/transformers/modeling_utils.py#L6186)? Any concerns?