[Agent] Implement system I/O collector (Network & Disk)

[varev-dev]

Goal

Implement an agent collector that gathers network and disk I/O metrics on Linux at configurable intervals and publishes them through the existing metric pipeline.

Acceptance criteria

• Collector starts with the agent and can be enabled in default official plugin set.
• Emitted metrics include (or equivalent): network.bytes.sent/received, network.packets.sent/received, network.errors.in/out.
• Emitted metrics include: disk.bytes.read/written, disk.ops.read/write.
• disk.utilization.percent is emitted (if available from source data).
• Collector handles missing/unavailable files gracefully (no agent crash; warning log + continue).
• Metric units are consistent and documented (bytes, ops, percent, per_second where applicable).

Sub-tasks & Scope of work

• Data Acquisition: Create collector module reading from /proc/net/dev and /proc/diskstats with robust parsing and clear error handling.
• Rate Computation: Compute rate metrics from monotonic counters between sampling ticks (delta/time).
• Labeling: Add low-cardinality labels (interface, device); skip loopback/virtual devices by default.
• Pipeline Integration: Integrate with ring buffer write path and ensure compatibility with variable collector intervals. Expose metrics in both push stream and local /metrics shim output.
• Testing: Write unit tests for parsing + rate calculations, and integration tests for collector registration/output shape.
• Documentation: Add short docs describing metric names, labels, units, known Linux limitations, and example config snippet.