fix: handle QuickTime wave/esds in mp4a codec derivation and fragmented init

[lideen]

Description

I noticed some QT files were producing incomplete/incorrect audio codec strings and compatibility issues in my fragmented playback flows.

This PR improves handling of QuickTime-style audio sample entries in QT files where codec/config metadata is nested under wave.esds rather than exposed directly on the sample entry.

The underlying issue was about how mp4a sample entries can be structured in QuickTime files: mp4a.wave.esds instead of mp4a.esds.

This has two main issues:

1. Codec derivation path for mp4a relies on direct esds
   • Nested wave.esds layouts fails to give the correct codec string.
2. The current init segment generation does not work for nested wave.esds. The MSE apis expects init segments to have direct esds.

What changed

1. mp4a codec derivation now supports nested QuickTime layout
   • For mp4a entries, codec derivation now resolves descriptor info from:
     • direct esds
     • nested wave.esds
2. Init segment writing now normalizes mp4a.wave.esds layouts
   • During writeInitializationSegment(...), mp4a entries with nested wave.esds are normalized to direct esds representation.
   • Original in-memory parsed structures are restored after writing (no persistent source mutation).