Harden CI against flaky Chromium install and ensure tests run reliably (child of #579)

[kalwalt]

Summary

A recent CI run failed during apt installation of chromium-browser (snap-backed) due to a Snapcraft assertion fetch timeout (HTTP 408). The job later passed on rerun, indicating a transient upstream/network failure — but the failure exposed fragility in our CI provisioning of headless Chromium.

Goal

Harden CI so test runs are not blocked by transient snap/network failures and ensure the refreshed test suite in #579 runs reliably in CI.

Proposed tasks

• Replace apt-based chromium install with a Playwright-managed Chromium binary (npx playwright install --with-deps chromium) or another stable binary provider.
• Add a short retry wrapper around package installation steps as a temporary mitigation.
• If immediate fix is needed, consider switching runner to ubuntu-22.04 as a short-term workaround.
• Update CI docs / README with instructions for running tests locally (how to get the browser binary).
• Add a smoke test step that runs a minimal NFT load + process() to catch infra-level failures early.

Logs / evidence

• Failing run: https://github.com/webarkit/jsartoolkitNFT/actions/runs/28015483901/job/82953947673
• Error snippet from runner logs:
  • "cannot fetch assertion ... unexpected HTTP status code 408"
  • "dpkg: error processing archive /var/cache/apt/archives/chromium-browser_..."
  • "E: Sub-process /usr/bin/dpkg returned an error code (1)"

Relation to #579

The work to refresh the tests should be tracked under #579. This issue is a CI/infra action item to make sure the refreshed tests run reliably in CI. Make this issue a child of #579.

Implementation notes

• Prefer Playwright for controlled browser binaries; add devDependency on playwright and set CHROME_BIN from require('playwright').chromium.executablePath() in the workflow.
• Keep a short-term retry loop if replacing the installation is not possible immediately.