feat: add per-track segmentation init mode

[lideen]

Summary

This PR adds a new segmentation initialization mode to support one init segment per fragmented track while preserving the existing combined-init behavior.

The main motivation is Media Source Extensions (MSE) integration: playback setups that require simultaneous multi-audio playback need one MediaSource (containing its SourceBuffer) per track, which in turn requires one init segment per track instead of a single combined init segment.

Note

This behavior aligns with historical usage, per-track initialization used to be the default behavior.
It stopped being the default in commit 77247e1 (fix: segmenting will also multiplex the fragmented tracks, Jul 8, 2025), first released in v1.3.1.

initializeSegmentation now supports:

• initializeSegmentation() / initializeSegmentation('combined') (default):
  returns a single init segment containing all selected fragmented tracks.
• initializeSegmentation('per-track'):
  returns one init segment per selected fragmented track.

Documentation

Updated README.md segmentation docs (it seems to have been stale documenting the old behavior before 77247e1):

• documented initializeSegmentation(mode) with supported values ('combined' and 'per-track')
• updated examples to reflect current return shapes
• fixed the setSegmentOptions anchor link