Bug: Fix critical audio processing bugs

[Leostrange]

Critical bugs found in code analysis

Bug 1 — SegmentSlicer: hardcoded MP3 format (ignores user selection)

File: app/src/main/java/com/example/vocalremover/data/SegmentSlicer.kt
Segment output filenames are hardcoded to .mp3 regardless of user-selected format in ProcessingConfig.
Fix: Use config.format.extension instead of hardcoded "mp3".

Bug 2 — WaveformCanvas: wrong segment time-to-X calculation

File: app/src/main/java/com/example/vocalremover/ui/components/WaveformCanvas.kt
The formula (segment.start / (duration / 1000f)) * size.width is incorrect — segments are in seconds, duration in milliseconds.
Fix: Use (segment.start * 1000f / duration) * size.width

Bug 3 — ResultsScreen: segments never passed from MainViewModel

File: app/src/main/java/com/example/vocalremover/ui/screens/ResultsScreen.kt
segments: List<Segment> = emptyList() — always empty, waveform never shows vocal regions.
Fix: Pass segments from ViewModel state through Screen.Results and into ResultsScreen.

Bug 4 — MediaExtractorWaveformGenerator: incorrect signed PCM parsing

File: app/src/main/java/com/example/vocalremover/data/waveform/MediaExtractorWaveformGenerator.kt
Manual bit manipulation for 16-bit PCM is error-prone. Use ByteBuffer.getShort() instead.

Acceptance Criteria

• Segments export in the format selected by user (MP3/WAV/FLAC/OGG)
• Waveform highlights vocal regions correctly
• Segments are visible on waveform in ResultsScreen
• Waveform amplitudes are rendered correctly