Use FrameOutput for audio as well

[NicolasHug]

See comment below

[NicolasHug]

Driveby: removing unused streamIndex field

[NicolasHug]

So this PR is about not addressing this TODO. I tried to address it, but then realized that we do call convertAVFrameToFrameOutput() in getNextFrameInternal():

torchcodec/src/torchcodec/decoders/_core/VideoDecoder.cpp

Lines 583 to 589 in b830480

	VideoDecoder::FrameOutput VideoDecoder::getNextFrameInternal(
	std::optional<torch::Tensor> preAllocatedOutputTensor) {
	validateActiveStream();
	UniqueAVFrame avFrame = decodeAVFrame(
	[this](const UniqueAVFrame& avFrame) { return avFrame->pts >= cursor_; });
	return convertAVFrameToFrameOutput(avFrame, preAllocatedOutputTensor);
	}

If we were to address the TODO, we would need separated audio/video logics in getNextFrameInternal(): one calling convertAVFrameToFrameOutput() (for video), and the other calling the audio-conversion stuff.

And we can't get around the fact that getNextFrameInternal() returns a FrameOutput, can we?

So, I guess this PR is about accepting FrameOutput as a "universal frame output", for both audio and videos.