[Feature] Rigorous copilot testing framework and domain-organized tool hierarchy

[pskeshu]

Description

Establish rigorous testing for copilot intelligence and reorganize tools into a proper microscopy domain model.

Current State

• 40+ tools registered via decorators
• Basic category organization (ACQUISITION, MOVEMENT, etc.)
• No systematic testing of copilot behavior

Requirements

1. Tool Domain Model

Reorganize tools into microscopy-specific domains:

tools/
├── sample/           # Sample operations (embryos, wells, dishes)
│   ├── detection.py
│   ├── tracking.py
│   └── state.py
├── acquisition/      # Image acquisition
│   ├── volume.py
│   ├── timelapse.py
│   └── snapshot.py
├── hardware/         # Direct hardware control
│   ├── stage.py
│   ├── optics.py
│   └── illumination.py
├── analysis/         # Image analysis
│   ├── focus.py
│   ├── detection.py
│   └── measurement.py
└── experiment/       # Experiment management
    ├── session.py
    └── workflow.py

2. Copilot Testing Framework

• Scenario-based tests with expected tool call sequences
• Mock hardware for offline testing
• Evaluation metrics: task completion, tool selection accuracy
• Regression test suite for known behaviors

3. Tool Expansion

• Gap analysis for missing microscopy operations
• Compound tools for common workflows
• Better tool documentation for Claude context

Technical Approach

• Refactor gently/agent/tools/ directory structure
• Create tests/copilot/ with scenario tests
• Use pytest with async support
• Mock QueueServerClient for offline tests

Key Files

• gently/agent/tool_registry.py
• gently/agent/tools/*.py
• New: tests/copilot/