pub fn start_recording(world: &mut World, max_seconds: f32) -> RecordingHandleExpand description
Start recording everything world fills, for at most max_seconds.
The recorder runs after every other OnAudioFill source and keeps the mix after
the world’s GlobalAudioGain and the unit-range clamp, so the result is what
the output stream receives. Because a source is handed the buffer before
the dispatcher scales it, the recorder applies the gain itself; that is the
one step it repeats rather than observes. No call fills the world: whichever
driver usually does (the audio thread, dispatch_audio_fill directly, a
stepped world) keeps doing it, and the recording grows as it goes.