Stable Segmentation, Smarter Playback Highlighting and Preserved Zoom

🔄 Improvements

• Reduced memory usage during live recording (lower RAM usage, fewer stalls on long sessions)
• Segmentation now applies server cuts more reliably and immediately when valid (more stable structure during long recordings)
• Improved handling of short transcription buffers (continuous auto-splitting without edge-case breakdowns)
• Enhanced header resize handling to prevent layout inconsistencies
• Improved dictaphone panel menu behavior based on settings (cleaner interface)

✨ New Features

• Highlight transcript words and phrases during playback (easier to follow, less rewinding)
  https://youtu.be/b8Bm_ccbJfI

• Preserve video and slide zoom when opening the next branch (no need to readjust scale every time)
  https://youtu.be/6a9Is2Q30nc

🛠 Fixes

• Fixed: Autoscroll is now disabled when no recording is active (no unexpected jumps, easier navigation)
• Fixed: Segmentation freeze on short buffers after long capture sessions (stable auto-splitting)
• Fixed: Forced segmentation fallback now always evaluates a valid cut range instead of stalling
• Fixed: Header size now updates correctly when content changes (no geometry issues)
  Before: https://shottr.cc/s/1tCo/SCR-20260226-hn2.png

And other improvements