epic(video): normalize video owned asset patterns to noun join tables

[willgriffin]

Related standardization epic: #1063

Summary

Evaluate and normalize smrt-video owned asset modeling onto the same noun-join pattern used elsewhere.

smrt-video already models owned assets, but it currently does so through STI Asset subclasses such as CharacterAsset, PerformerAsset, SceneAsset, VideoShotAsset, VideoSequenceAsset, and VideoCompositionAsset.

Why

• the current pattern is consistent in intent but different in physical storage/modeling from content_assets
• if we want one repo-wide primitive asset ownership pattern, video should converge too
• this is a larger migration than the primitive join-table additions and should be tracked separately

Scope

• evaluate whether to replace STI Asset subclasses with noun join tables for:
  • characters
  • performers
  • scenes
  • video shots
  • video sequences
  • video compositions
• preserve role typing and existing behavior during migration
• document migration path and compatibility considerations

Out of scope

• changing generic/provenance AssetAssociation semantics
• folding video into generic asset associations

Done when

• we have a clear decision and implementation path for video asset normalization