feat(video): implement M7 video generation, ComfyUI macros, cloud calling, and canvas video cards

[BerryUIKI]

Summary

Delivers Milestone M7: Video generation and usable video creation workflow in Berry AI Studio:

• Schemas & Protocols: Added TXT2VIDEO and IMG2VIDEO to CreativeActionType, extended CreativeActionRequest, GenerationProvenance, and CreativeActionResult with video parameters (ps,
  um_frames, motion_bucket_id, duration_seconds).
• Deterministic Caching: Integrated video-specific parameters into compute_creative_cache_hash ensuring deterministic dirty-checking.
• ComfyUI Video Macros: Added �uild_comfy_img2video_graph (Stable Video Diffusion XT) and �uild_comfy_txt2video_graph (AnimateDiff) in �ackend/app/runners/macro_compiler.py.
• Cloud BYOK Video: Added _call_fal_ai_video (Fast SVD for img2video, Luma Dream Machine for txt2video) and _call_siliconflow_video (CogVideoX) in �ackend/app/runners/api_runner.py.
• Asset Storage: Added save_media_from_url in AssetStore supporting MP4 and WebM video media types.
• Frontend Video Cards & Controls:
  • ImageCardNode: Renders interactive
  • VideoModal: Dedicated modal for parameterizing image animation (FPS: 8/16/24, frames: 16/25, motion bucket ID: 1..255, engine selector).
  • CreationDock: Added direct Video creation action.
• Automated Tests: Added �ackend/tests/test_m7_video.py covering schemas, cache hashing, ComfyUI macro graphs, cloud video execution, and error handling (80/80 backend tests passing).

Test Plan

• Ran backend test suite: 80 passed in 37.31s.
• Ran launcher tests: cargo test passed.
• Ran frontend build: tsc && vite build passed with 0 errors.