Enhance pet clustering and identity workflows with error handling

[OliverZhaohaibin]

This pull request introduces a comprehensive Pets recognition feature to the application, integrating pet detection, clustering, and state management alongside the existing People recognition system. The documentation has been updated throughout to reflect these architectural and functional changes, including details on the new Pets runtime, its interaction with People, and the new database/state artifacts. Startup and worker lifecycles are also improved to coordinate scanning and AI workloads more robustly. Below are the most important changes:

New Pets Recognition Feature:

• Added an optional Pets pipeline using YOLOX/ONNXRuntime for cat/dog detection and DINOv2 for pet identity embeddings, with species-separated clustering and robust handling for missing dependencies or models. [1] [2]
• Introduced independent, library-scoped Pets state, including new database files (pet_index.db, pet_state.db), session-bound services, and UI integration for pet cards, naming, merging, hiding, and group composition with People. [1] [2] [3] [4]

Architecture and State Management Updates:

• Updated the architecture documentation and diagrams to reflect the new Pets bounded context, ports, and runtime/session composition, ensuring Pets and People remain independent but composable. [1] [2] [3] [4] [5]
• Clarified session and service binding rules to prevent fallback or duplicate service creation outside of the active library session.

Startup and Worker Coordination:

• Deferred metadata scanning and AI worker startup until the gallery is warmed, with fallback timers; coordinated scan workers, face, and pet AI workers to operate independently and robustly during rescans and shutdown. [1] [2] [3] [4]

User Interface and Interaction Improvements:

• Extended the dashboard and overlays to display pet annotations, allow pet identity management, and support mixed People & Pets groups, while ensuring cross-type detection conflicts are resolved gracefully. [1] [2] [3]

Documentation and Developer Guidance:

• Added and referenced docs/misc/PETS_RECOGNITION_RUNTIME.md for maintenance and implementation details, and updated installation, packaging, and dependency notes for the new recognition runtimes. [1] [2] [3]

These changes significantly enhance the application's recognition capabilities and modularity, while maintaining clear separation and graceful degradation for optional AI features.