MVP-2: Deploy Redis for task queue and pub/sub

## Problem
No Redis instance in cluster for:
- Task queue (Celery/RQ)
- Pub/Sub for real-time monitoring
- State persistence for orchestrator
- Session management for agents

## Requirements
1. Deploy Redis StatefulSet
2. Create PersistentVolumeClaim for data
3. Configure Service for internal access
4. Add health checks
5. Configure Redis for orchestrator use cases

## Implementation Plan
1. Create Redis StatefulSet manifest
2. Create PVC for data persistence
3. Create Service for internal access
4. Add resource limits
5. Test connectivity from orchestrator

## Acceptance Criteria
- Redis is running and accessible
- Data persists across pod restarts
- Health checks pass
- Orchestrator can connect and use Redis

Start by locating the Kubernetes deployment manifests and the orchestrator’s Redis connection configuration. Review the StatefulSet, PVC, Service, resource limits, and health-check requirements, then test connectivity and persistence across pod restarts. Done means Redis is accessible to the orchestrator, survives pod restarts, and passes health checks.

Written by the indexing model from the issue text.