Add single-source split/delete timeline editing

[brunomendonca-com]

Summary

Add single-source, non-destructive split/delete editing while keeping export as one MP4:

• T splits at the playhead and selects the right-hand clip.
• Each clip has independently draggable start/end handles. Retained ranges remain in source order and cannot overlap.
• Delete / Backspace removes the selected clip and closes the gap.
• Split/Delete buttons include shortcut tooltips and accessible names; help and README are updated.
• Preview and seeking traverse retained ranges. Export concatenates those ranges with ffmpeg, supporting sources with or without audio.
• Deleting the final clip clears the video/timeline and returns to Open a video.

The implementation includes a testable clip model, whole-timeline dirty-state tracking, export snapshots, and file-dialog guards/cancellation handling. Multi-range export belongs with split editing because independently resizing split clips can already create gaps.

Scope / draft status

This is a proposal to extend the current single-range trimmer. Keeping it draft for feedback on whether single-source multi-clip editing fits Omacut's scope, and for further desktop/media validation.

No undo, reordering, transitions, multiple source files, or project persistence. Preview uses source seeks and is not guaranteed seamless across cuts. The timeline currently uses a representative thumbnail per clip.

Validation

On the standalone branch:

• QT_QPA_PLATFORM=offscreen ./bin/test: 36 backend/UI/export + 11 clip-model passes, including QtTest initialization/cleanup.
• ./bin/build: successful.
• git diff --check: clean.

Tests cover split/resize/delete boundaries, both deletion keys and buttons, independent handle dragging, actual preview jumps over removed material, export snapshot isolation, invalid ranges, final-clip reset/reloading, and synthetic color-sequence exports verifying duration/order with and without audio.

Remaining validation: interactive desktop testing, frame-rate variations, unusual audio offsets/synchronization at cuts, and more real-world exports. A desktop-platform run of the combined development branch failed the right-handle mouse-drag test while offscreen passed; that discrepancy still needs investigation. Existing scale-argument tests pass, but broader multi-range scaling integration coverage is still needed.

Independence

This PR targets upstream master directly and has no dependency on an unmerged PR. Number-key seeking is deliberately excluded and proposed separately against the existing upstream trim API. If that proposal merges first, this PR will adapt it to retained timeline time during integration.

[tristonarmstrong]

you should include a demo with this pr