FEAT: Support audio model

[codingl2k1]

• Currently, only whisper model is supported.
• Provide /v1/audio/transcriptions and /v1/audio/translations.
• Compatible with openai API.

Example:

import openai

client = openai.Client(api_key="not empty", base_url=f"{endpoint}/v1")
with open(zh_cn_audio_path, "rb") as f:
    completion = client.audio.transcriptions.create(model=model_uid, file=f)
    assert "列表" in completion.text
    assert "香港" in completion.text
    assert "航空" in completion.text

    completion = client.audio.translations.create(model=model_uid, file=f)
    translation = completion.text.lower()
    assert "list" in translation
    assert "airlines" in translation
    assert "hong kong" in translation