Handle video-only or audio-only streams in segmenter

Consider these cases:

* Audio or video only streams
* Streams that come in late; eg past analyzeduration
* Invalid video, such as video that doesn't have keyframes. It may still be useful to write audio alone to the segment. Vice versa for audio.

No file or test is named. Start by locating the segmenter entry point and its existing tests, then reproduce audio-only, video-only, late-arriving, and keyframe-less inputs. Done means the segmenter preserves usable audio or video when the other stream is absent, late, or invalid.

Written by the indexing model from the issue text.