Add Bot Manager for Bot Lifecycle Coordination

[tgrunnagle]

Closes #98

Summary

Implements BotManager, the central coordinator for bot lifecycle management in the Bot Service. This component bridges the REST API layer with bot runners and WebSocket clients, enabling spawning bots into game rooms, loading trained neural network models from the ModelRegistry, and managing multiple concurrent bot instances.

Changes Made

bot2/

• Added BotManager class in bot/service/bot_manager.py with full lifecycle management:
  • spawn_bot() creates bots and connects them in background tasks
  • destroy_bot() gracefully stops and removes bots
  • get_bot() and list_bots() for status queries
  • list_models() returns available models from registry
  • shutdown() for graceful service termination
• Added Pydantic models: BotConfig, SpawnBotRequest, SpawnBotResponse, BotInfo
• Updated bot runner constructors to accept optional client parameter (set later by WebSocketBotClient)
• Exported new types from bot/service/__init__.py

Implementation Details

• Bot IDs use human-readable format: bot_{uuid_hex[:12]}
• Background connection via asyncio.create_task() allows immediate API response while bot connects
• BotConfig validates that rule-based bots cannot specify model parameters, and neural network bots must provide either model_id or generation
• Failed background connections are automatically cleaned up from the bot registry

Testing

• Added 1000+ lines of unit tests covering:
  • Initialization with and without registry
  • Spawning rule-based and neural network bots
  • Model loading by ID and by generation
  • Error handling (missing model, no registry, invalid config)
  • Unique bot ID generation
  • Bot destruction and status queries
  • Graceful shutdown
  • Pydantic model validation
• Tests use mocking to avoid actual network connections and model loading