🐳 Optimize Docker image size

Description

Create a multi-stage Dockerfile to reduce the final image size and improve deployment efficiency.

Acceptance Criteria

• Create optimized multi-stage Dockerfile
• Use Alpine or distroless base image for final stage
• Final image should be under 50MB
• Include only necessary runtime dependencies
• Add .dockerignore file to exclude unnecessary files
• Update README with Docker usage instructions

Current Advantages

The switch to built-in models provides significant advantages for Docker optimization:

• ✅ No external model files to copy (models compiled into binary)
• ✅ Smaller attack surface (no external dependencies)
• ✅ Faster startup (no model loading from disk)
• ✅ Single binary deployment

Implementation Guidance

• Use rust:alpine for build stage
• Use scratch or alpine for runtime stage
• Copy only the furnace binary (no model files needed)
• Consider static linking for zero dependencies
• Test with --model-name resnet18 (built-in model)
• Update .dockerignore to exclude models/ directory

Expected Image Size

With built-in models, the final image should be significantly smaller:

• Binary size: ~4.5MB
• Base image: ~5MB (Alpine) or ~2MB (distroless)
• Target: Under 20MB (more aggressive than original 50MB target)

Estimated Difficulty

Easy - Half day (easier due to built-in models)