fix: stream artifact downloads to disk to avoid OOM on large artifacts

[yeicor]

In main.js, every artifact ZIP is currently downloaded by calling Octokit's downloadArtifact, which follows the redirect and fetches the entire binary response body into a zip.data ArrayBuffer that lives in Node.js heap memory. That Buffer.from(zip.data) is then passed either to AdmZip (which also keeps everything in RAM) or written to a temp file for unzip. For large artifacts this can easily exhaust the runner's memory and kill the process with an OOM error.

This PR fixes this issue by streaming the downloads to disk.

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Streams GitHub Actions artifact ZIP downloads directly to disk to avoid OOM when processing large artifacts.

Changes:

• Added downloadArtifactToFile() to stream ZIP responses to a file via a custom Octokit fetch implementation.
• Updated artifact download/extraction flow to use temp ZIP files and clean them up afterward.
• Adjusted error handling for expired artifacts and extraction paths.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[yeicor]

CI passes on my fork: https://github.com/yeicor/action-download-artifact/actions

Fixes #201

[dawidd6]

Could you please address the copilot review comments?

[yeicor]

All Copilot comments addressed with minimal changes. CI passes: https://github.com/yeicor/action-download-artifact/actions/runs/25122241313

• Log group closure fixed using inner try/finally to always call core.endGroup().
• Redirect explicitly set to "follow" to avoid relying on defaults.
• Runtime already satisfies Node 18+ via node24 (added a comment about this).

Note: AdmZip still loads full ZIP into memory; use_unzip: true is required for streaming extraction (maybe this should be documented somewhere or be the new default).

[dawidd6]

A bit too much changes for just a change making the artifact download streamable. Would be nice to have the download + extract logic both streamable. GitHub does so in actions/toolkit using unzip-stream package. Maybe we could do that too.

[yeicor]

Thanks for the feedback. This is currently good enough for me, and I don’t really have time to keep iterating on it. Happy for you (or anyone) to take this branch and build on it, though.