v0.6.0: Multi-core Parallel Processing

What's New

• Parallel layer downloads: Utilizes worker pool with auto-detected CPU cores (max 8 workers)
• Parallel compression: Uses pgzip for multi-core gzip compression
• Streaming memory: Processes up to 4 layers concurrently while maintaining low memory footprint
• Performance improvement: 3-5x faster downloads, 2-4x faster compression for large images

Performance

For large images like vllm (22+ layers):

• Multi-core CPU utilization instead of single-core bottleneck
• Significantly reduced export time for first-time downloads
• Improved throughput when using layer cache

Technical Details

• New worker pool implementation in internal/image/layer_processor.go
• Ordered buffer ensures correct layer sequence for tar format
• Context-aware for proper cancellation support

Full changelog: v0.5.2...v0.6.0