feat(content-studio): compose scene clips into final vertical video export

[thxforall]

Parent: #508
Depends on: #519, #522

Context

Phase 1A/1B generates and stores scene-level clips, but Content Studio still needs a server-side composition stage that turns those clips into one finished short-form asset.

Scope

• Add ffmpeg-based clip stitching/composition for generated scene clips.
• Produce a final 9:16 vertical export by default.
• Add subtitle/text overlay support using the ShortFormPlan voiceover/script data.
• Add BGM and voiceover audio track support.
• Store and expose the final composed video asset for preview/download.
• Keep failures attributable to either scene generation or final composition.

Acceptance Criteria

• A job with completed scene clips can produce one final downloadable 9:16 video.
• Composition failures are visible separately from provider scene failures.
• Focused tests or smoke coverage validate ffmpeg command construction and final asset persistence without requiring paid provider calls.

Out of Scope

• provider-specific generation modes.
• governance and cost-tracking integration.
• advanced timeline editing UI.