Add restore network telemetry

[sjmiller609]

Summary

• adds nested OpenTelemetry spans for restore-time network work, so allocation, snapshot network preparation, TAP creation, bridge attachment, isolation, and rate-limit setup show up separately in traces
• carries inherited hypervisor trace attributes into network-manager spans so restore traces can be correlated across instance restore and network setup
• keeps the change telemetry-only: allocation, TAP, bridge, rate-limit, and restore control flow behavior is unchanged

Tests

• git diff --check origin/main...HEAD
• go test ./lib/network -count=1
• go test -tags containers_image_openpgp ./lib/instances -run TestCreateInstanceClearsRetentionStateBeforeMetadataSave -count=1

[firetiger-agent]

Monitoring Plan: OTel Tracing for Network Allocation and TAP Creation

What this PR does: Adds OpenTelemetry tracing instrumentation to the hypeman network allocation layer — covering IP/MAC/TAP creation and standby restore network setup. Improves debuggability of network failures with no behavioral changes.

Intended effect:

• New network.* and restore_network.* spans: baseline 0 spans visible (none existed pre-deploy); confirmed if spans named network.get_default_network, network.create_tap, restore_network.create_allocation appear in the trace pipeline post-deploy.
• Instance spawn rate: baseline 330K–835K/hr active hours; confirmed if rate remains within this range after deploy.

Risks:

• Context deadline propagation — if span contexts inadvertently shorten timeouts, TAP device creation could fail; alert if kernel_invocation_spawn_total drops below 300K/hr during active hours (Mon 09:00–19:00 UTC).
• Network allocation errors — failed to allocate network / failed to recreate network ERROR logs on hypeman nodes; alert if any sustained increase appears post-deploy (baseline: negligible).
• API error rate regression — API ERROR log count; alert if sustained above 80K/hr (baseline 47K–52K/hr).

Status updates will be posted automatically on this PR as monitoring progresses.

View monitor