More Stable Transcript Processing and Audio Playback

🔄 Improvements

• Improved audio stream reliability to prevent browsers from reusing broken cached audio responses after stream updates
• Improved consistency between browser playback and audio export handling
• Improved pasted transcript processing performance so large transcript imports no longer block the app

🛠 Fixes

• Fixed: browsers could continue using broken cached audio streams after stream updates, causing playback/export inconsistencies
• Fixed: long pasted transcripts could freeze processing or create empty dictaphone sections
• Fixed: generated dictaphone sections could lose transcription links during map/transcription processing
• Fixed: duplicate processing actions during large transcript imports could trigger unstable behaviour
• Fixed: pasted transcript imports now remain responsive during segmentation and AI processing Screenshot

And other improvements