Use raw format for AzureTTSService output

[amacapri]

In the AzureTTSService class, there is a comment that states:

# Azure always sends a 44-byte header. Strip it off.
yield AudioRawFrame(
        audio=result.audio_data[44:],
        sample_rate=self._sample_rate,
        num_channels=1,
    )

I believe this behavior occurs because the default speech_synthesis_output_format is being used, which is most likely in WAV format. Instead, you should consider using one of the "raw" formats, such as SpeechSynthesisOutputFormat.Raw16Khz16BitMonoPcm.