Refactor C++ SingleStreamDecoder to consolidate addStream into constructor

[scotts]

Right now, the C++ SingleStreamDecoder class has two constructors:

torchcodec/src/torchcodec/_core/SingleStreamDecoder.h

Lines 34 to 45 in 1ea235a

	// Creates a SingleStreamDecoder from the video at videoFilePath.
	explicit SingleStreamDecoder(
	const std::string& videoFilePath,
	SeekMode seekMode = SeekMode::exact);
	// Creates a SingleStreamDecoder using the provided AVIOContext inside the
	// AVIOContextHolder. The AVIOContextHolder is the base class, and the
	// derived class will have specialized how the custom read, seek and writes
	// work.
	explicit SingleStreamDecoder(
	std::unique_ptr<AVIOContextHolder> context,
	SeekMode seekMode = SeekMode::exact);

And separate from there, there is a public API for adding streams of different media types:

torchcodec/src/torchcodec/_core/SingleStreamDecoder.h

Lines 90 to 97 in 1ea235a

	void addVideoStream(
	int streamIndex,
	std::vector<Transform*>& transforms,
	const VideoStreamOptions& videoStreamOptions = VideoStreamOptions(),
	std::optional<FrameMappings> customFrameMappings = std::nullopt);
	void addAudioStream(
	int streamIndex,
	const AudioStreamOptions& audioStreamOptions = AudioStreamOptions());

Note that there is also a private member function that both of those call:

torchcodec/src/torchcodec/_core/SingleStreamDecoder.h

Lines 304 to 309 in 1ea235a

	void addStream(
	int streamIndex,
	AVMediaType mediaType,
	const torch::Device& device = torch::kCPU,
	const std::string_view deviceVariant = "ffmpeg",
	std::optional<int> ffmpegThreadCount = std::nullopt);

This separation is a relic of when SingleStreamDecoder was trying to be more than just a single stream decoder. This API does not match the public Python API, and that fact causes some awkwardness, in particular with how we deal with custom frame mappings; see PR #1060 (comment) for more.

Note that one could imagine the current separation enables just getting metadata without adding a decoding stream, but that should be well supported with SeekMode::approximate.

[scotts]

I realized that we can do this in stages:

1. Consolidate all of the core APIs related to video decoder creation and adding a stream. We don't need to actually change the C++ code to do this.
2. Then change the C++ code.