Update pipeline tag to audio-text-to-text

This PR updates the `pipeline_tag` for the model card from `audio-to-audio` to `audio-text-to-text`.

The model is described as an "Interleaved Speech-Text Language Model" that can generate "speech or text continuations over discrete Hubert tokens given speech-text prompts." This indicates that it processes both speech and text as input and can generate both speech (via vocoding Hubert tokens) and text as output. The `audio-text-to-text` pipeline tag accurately reflects this multi-modal input and output capability, improving the model's discoverability and categorization on the Hugging Face Hub.

README.md

@@ -1,13 +1,13 @@
 ---
-library_name: transformers
-license: llama3.2
+base_model:
+- meta-llama/Llama-3.2-3B
 datasets:
 - slprl/sTinyStories
 language:
 - en
-base_model:
-- meta-llama/Llama-3.2-3B
-pipeline_tag: audio-to-audio
+library_name: transformers
+license: llama3.2
+pipeline_tag: audio-text-to-text
 ---
 
 # Scaling Analysis of Interleaved Speech-Text Language Models