[P2] Add reliability and performance observability for import, render, and restore

[nkanf-dev]

Summary

Introduce lightweight observability for critical reliability and performance paths so regressions can be detected and prioritized with evidence.

Why

Without basic telemetry or structured diagnostics, performance and reliability work is largely guesswork.

Current evidence

• Import, thumbnail rendering, and restore flows currently rely on sparse console logging.
• There is no consistent record of latency, failure rate, or recovery outcomes.

Scope

• Define a minimal event set for import start/success/failure, thumbnail render success/failure, and restore success/failure.
• Capture duration where useful.
• Keep the mechanism lightweight and privacy-conscious.
• Document how developers can inspect these signals locally and in CI/dev environments.

Acceptance criteria

• Critical flows emit structured diagnostics consistently.
• Developers can inspect failure causes and rough latency without reproducing everything manually.
• Observability does not materially degrade runtime performance.

Out of scope

• Full analytics platform integration.