whisper openai.AudioRequest.FilePath

[xoba]

i would like to create whisper model transcriptions from audio in a memory-based buffer, ideally of type []byte; whereas currently the only way to identify audio in this sdk is via openai.AudioRequest.FilePath, thus it must be stored in the underlying OS's filesystem.

in my current use case, the audio file is already in a []byte buffer, never came from a filesystem in the first place, thus it feels unnecessary and artificial to explicitly store it in a temporary file and thereafter delete it in order to use whisper api via this projects sdk.

thanks!

[vvatanabe]

@xoba Done #331