Fix critical in-memory task storage with database persistence

Major architectural improvement to replace in-memory task storage with
database-backed persistence while maintaining backward compatibility.

Changes:
- Created Task SQLAlchemy model matching database schema
- Added Workflow and Execution SQLAlchemy models
- Created TaskService for database CRUD operations
- Updated UnifiedCoordinator to use database persistence
- Modified task APIs to leverage database storage
- Added task loading from database on coordinator initialization
- Implemented status change persistence during task execution
- Enhanced task cleanup with database support
- Added comprehensive task statistics from database

Benefits:
- Tasks persist across application restarts
- Better scalability and reliability
- Historical task data retention
- Comprehensive task filtering and querying
- Maintains in-memory cache for performance

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

backend/app/models/agent.py

@@ -1,5 +1,6 @@
 from sqlalchemy import Column, Integer, String, DateTime, JSON
 from sqlalchemy.sql import func
+from sqlalchemy.orm import relationship
 from ..core.database import Base
 
 class Agent(Base):
@@ -23,6 +24,9 @@ class Agent(Base):
     updated_at = Column(DateTime(timezone=True), onupdate=func.now())
     last_seen = Column(DateTime(timezone=True), nullable=True)
     
+    # Relationships
+    tasks = relationship("Task", back_populates="assigned_agent")
+    
     def to_dict(self):
         return {
             "id": self.id,