Add autoenroll and UUID

[YuchengZhou821]

This pull request introduces tier-aware identity tracking and a new sample refresh manager to improve the robustness and introspection of the face recognition system. The main changes include adding a SampleRefreshManager for controlled gallery updates, updating WhoTracker to track confidence tiers for identities, and enhancing the snapshot diagnostics with tier breakdowns and improved windowed statistics.

Tier-aware identity tracking and diagnostics:

• who_tracker.py:
  • The WhoTracker class now tracks not just names but also confidence tiers ("confident", "tentative", "uncertain") for each identified face. This enables more granular downstream behavior (e.g., robot greeting only on "confident" matches) and richer diagnostics.
  • The snapshot method now includes per-frame and windowed breakdowns by tier (confident_now, tentative_now, confident_frames, tentative_frames), while preserving legacy fields for backward compatibility. [1] [2]
  • All event storage and processing is updated to handle parallel lists of names and tiers.

Sample refresh management:

• sample_refresh.py:
  • Introduces a new SampleRefreshManager class that manages when to add new samples to the face gallery. It enforces configurable guards for match confidence, per-UUID rate limiting, and diversity checks, preventing gallery pollution and resource overuse.

Module and dependency updates:

• __init__.py:
  • Adds "selfie_logic" to the module exports to ensure tier constants are kept in sync across modules.

These changes collectively make the face recognition pipeline more robust, introspectable, and ready for tier-aware downstream consumers.