Add to docs, how to end `StreamedAudioInput` gracefully

[majklost]

Describe the bug

In current docs, there is no information how the end of the input stream is signalized for StreamedAudioInput. After searching through sources, I found out that the correct solution is to pass None instead of numpy frame

As can be seen in

openai-agents-python/src/agents/voice/models/openai_stt.py

Line 254 in 329806a

if buffer is None:

I suggest updating description of StreamedAudioInput. Will do this in PR

Repro steps

Current docs do not have information about a way to terminate StreamedAudioInput

Expected behavior

A clear and concise description of what you expected to happen.

[seratch]

Fixed by #1687