MVP-5: Implement Agent Runner

## Problem
Need an agent runner that:
- Runs in ephemeral K8s Jobs
- Connects to Redis for task queue
- Connects to LiteLLM for model access
- Connects to GitHub for PR creation
- Handles errors and retries autonomously
- Logs structured data for monitoring

## Requirements
1. Create Agent Runner component
2. Implement Redis task queue consumer
3. Implement LiteLLM client for model routing
4. Implement GitHub integration for PR creation
5. Add error handling and retry logic
6. Add structured logging

## Implementation Plan
1. Create Agent Runner Python module
2. Implement Redis consumer
3. Implement LiteLLM client
4. Implement GitHub integration
5. Add error handling
6. Add logging

## Acceptance Criteria
- Agent Runner can consume tasks from Redis
- Agent Runner can create PRs on GitHub
- Agent Runner handles errors gracefully
- Agent Runner logs structured data

No files, tests, or entry points are named. Start by reviewing the repository structure and existing k3s or Kubernetes configuration, then map the Redis, LiteLLM, GitHub, retry, and logging boundaries. Done means the runner consumes Redis tasks, creates GitHub PRs, handles failures gracefully, and emits structured logs.

Written by the indexing model from the issue text.