GGUF: Fix llama 3 GGUF (#31358)
* Create push-important-models.yml
* llama3 support for GGUF
* fixup
* Update src/transformers/integrations/ggml.py
* fix pre-tokenizer
* fix
* fix
* fix
* fix
* fix
* fix
* address final comment
* handle special tokens + add tests
