Remove `soundfile` import

The `soundfile` import is not used in the standalone acoustic model example. This PR removes this redundant import.

README.md

@@ -70,7 +70,6 @@ To transcribe audio files the model can be used as a standalone acoustic model a
 ```python
  from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
  from datasets import load_dataset
- import soundfile as sf
  import torch
  
  # load model and tokenizer