Ability to decode readable stream

Buffering an audio file can be expensive in terms of memory, it'd be nice if the lib could decode the readable stream as well.

```ts
const stream = fs.createReadStream(file)
const decoded = decodeAudioStream(stream)
```

Perhaps symphonia can be customized to decode a chunk at a time?

Here's a pseudocode for example:
```ts
// Transform stream (node)
_transform(chunk, _, done) {
this.push(binding.decodeAudioChunk(chunk))
done()
}
```

```rs
fn decode_audio_chunk(chunk: Vec) {
...
}
```

Start by tracing the existing Symphonia decoding path and the Node binding entry point. Define how decodeAudioStream and decodeAudioChunk should handle chunk boundaries and completion without buffering the whole file. Done means a readable stream can be decoded incrementally with bounded memory and the resulting audio matches current decoding.

Written by the indexing model from the issue text.