fix: isolate exporter timeout phases

[jingxiang-z]

Summary

Fix exporter timeout handling so a stalled collection phase does not force HTTP upload to inherit an expired context.

What Changed

• split exporter collection and upload into separate timeout-scoped contexts
• stop collector work promptly when cancellation or deadline expiry is observed
• preserve partial event data collected before cancellation
• add regression tests for partial collection and fresh upload context behavior

Root Cause

The exporter previously used one shared timeout for metadata refresh, collection, and upload. When GPU-facing collection work stalled, the shared context expired and the later event-read and HTTP-upload phases inherited a dead context, causing cascading context deadline exceeded failures.

Impact

The agent now degrades more cleanly under driver/NVML stalls: partial collection can still be uploaded, and cancellation no longer produces misleading per-phase timeout cascades.

Validation

• go test ./internal/exporter/...

[coderabbitai]

Important

Review skipped

Draft detected.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 23c5a54e-f926-4d56-9c6b-82df12ca3aeb

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex-fix-exporter-timeout-cascade

---

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[jingxiang-z]

Superseded by #154, which uses the repo-conventional branch name fix/exporter-timeout-phases.