[Feature] Full detector system integration with async acquisition hooks

[pskeshu]

Description

Complete the integration between the detector system, pre/post-detection hooks, and the async acquisition pipeline via the sample state entity (embryos).

Current State

• DetectionQueue processes volumes through detectors
• Callback chain exists: volume → detection → state update
• Sample state (embryos) stores detection results
• Pre/post hooks not fully implemented

Requirements

1. Pre-Detection Hooks

• Image preprocessing before detector runs
• Configurable per detector
• Examples: contrast enhancement, ROI extraction, temporal averaging

2. Post-Detection Hooks

• Actions triggered after detection fires
• Per-embryo state updates
• Interval rule adjustments
• Notification/logging

3. Sample State Connection

• /embryos endpoint as central sample registry
• Detection results → embryo semantic state
• Acquisition parameters ← embryo requirements
• Bidirectional: detections inform acquisition, acquisition feeds detection

4. Hook Configuration

• Declarative hook definitions
• Chain multiple hooks
• Conditional execution

Technical Approach

• Extend Detector class with pre/post hook lists
• Create hook base class with execute(context) interface
• Connect to EmbryoState.detection_results for state updates
• Add semantic state fields to EmbryoState (developmental_stage, health_status)

Key Files

• gently/agent/detector.py
• gently/agent/detection_queue.py
• gently/agent/state.py (EmbryoState)
• gently/agent/timelapse_orchestrator.py