feat: add agent self-observability metrics

[mayankpande88]

Problem

The agent has zero visibility into its own health. Operators can't tell when data is being lost, buffers are full, or parsing is failing.

Proposed Metrics

Metric	Type	Description
node_agent_events_processed_total	Counter	Events processed by type (tcp, l7, proc, file)
node_agent_events_dropped_total	Counter	Events dropped (channel full, parse error)
node_agent_ebpf_lost_samples_total	Counter	Lost samples per perf/ring buffer
node_agent_l7_parse_errors_total	Counter	L7 parse failures by protocol
node_agent_uprobe_attach_total	Counter	Uprobe attach attempts (success/failure/skipped)
node_agent_event_channel_usage	Gauge	Current event channel buffer utilization
node_agent_pending_l7_events	Gauge	L7 events waiting for TCP connection match
node_agent_ip2fqdn_evictions_total	Counter	ip2fqdn map entries evicted at cap
node_agent_heap_bytes	Gauge	Go heap in-use bytes (from runtime.MemStats)
node_agent_goroutines	Gauge	Active goroutine count

Impact

Critical for operating a fleet — currently blind to data loss.