Add eBPF-based IO performance monitoring and diagnostic tool to resolve application request delays caused by full disk IO bandwidth or sudden IO access spikes.

Key Features:

• Monitor block layer and filesystem IO operations, collecting latency, throughput and other metrics
• Support multi-device filtering to focus on IO from specific devices
• Automatically trigger data collection when IO latency or utilization exceeds thresholds
• Provide context information including filename, path, disk device, inode number, container, etc.
• Collect IO scheduling latency call stacks to help identify root causes

Implementation Components:

• bpf/iotracing.c: Core BPF program implementing kernel-level IO data collection
• cmd/iotracing/: Command-line tool supporting multi-device filtering and multiple output formats
• core/autotracing/iotracing.go: Automatic triggering mechanism and threshold detection
• internal/conf/config.go: Configuration parameter definitions