fix(scripts): give each package a time budget

[noctuum]

An upstream host that accepts a connection and never answers held prepare or verify until the 30-minute job timeout. publish was then skipped, and no issue opened.

• check-all.sh runs the script of each package under timeout --kill-after=30: 120 s for the version check, 300 s for an update, verify or publish. The whole process group ends, and the error of the package is timed out after <n> s.
• The Hand over step waits up to 30 s, and only for live processes of builder. It hung on a zombie before, since pid 1 of the container never reaps one.
• publish runs after a failure of prepare or verify as well. It publishes nothing then and records update_failed for every package with a new version.

bats tests/ --recursive: 390 pass, actionlint is clean.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!