CI: retry ext-host remote tests on Electron startup segfault

[timtmok]

Summary

Since the Electron 42 bump (upstream 1.124.0 merge, #14673, landed July 8), the test / ext-host job's Run Extension Host Tests (Remote) step intermittently fails with exit code 139. Electron segfaults at startup, before any test runs, inside the system libfontconfig / libexpat on the ubuntu24 CI image:

Received signal 11 SI_KERNEL
#3 libexpat.so.1.9.1
#6-#11 libfontconfig.so.1.12.1
#12 libpangoft2-1.0.so
#13 libglib-2.0.so (worker thread)
./scripts/test-remote-integration.sh: line 91: Segmentation fault (core dumped) "$INTEGRATION_TEST_ELECTRON_PATH" ...

The crash strikes whichever suite happens to be launching Electron at the time - observed against the TypeScript, ipynb, and configuration-editing suites on unrelated branches (runs 29043333717, 29094973481, 29060813983, 29066567774, 29094468926). No ext-host failures with this signature exist before July 9; since then it has killed the majority of ext-host runs.

Fix

Retry the remote integration step up to 3 times, but only when the exit code is 139, so genuine test failures still fail fast on the first attempt. scripts/test-remote-integration.sh is upstream-owned, so the retry lives in the Positron-owned workflow instead.

This is a mitigation. The real fix is likely in the CI image (fontconfig/expat packages in positron-ci-images ubuntu24) or a future Electron release; the retry can be removed once startup is stable again.

QA Notes

CI-only change. Verify the test / ext-host job passes; on a segfault the step log will show Electron segfaulted at startup (exit 139) on attempt N/3 before retrying.

🤖 Generated with Claude Code

[github-actions]

E2E Tests 🚀
This PR will run tests tagged with: @:critical

^{readme  valid tags}