ci(release): stop and restart the daemon around cask upgrades

[madeinoz67]

Summary

brew upgrade gortex replaces the binary while a running daemon keeps executing the old image — nothing restarts it, and nothing detects the skew (#609). PR #611 adds the detection layers; this PR closes the loop at the install path: the release pipeline's generated cask now stops a running daemon before the upgrade and restarts it onto the new binary afterwards.

Fixes the brew-upgrade leg of #609 (detection: #611).

Changes

• .github/workflows/release.yml (the Publish homebrew cask heredoc template) gains guarded hooks in every generated Casks/gortex.rb:
  • pre_install: if a gortex binary exists AND gortex daemon status exits 0, write a tmp marker and run gortex daemon stop — releases the store lock and retires the old image before the swap.
  • post_install: if the marker exists, delete it and run gortex daemon restart (routes through the supervisor when launchd owns the daemon).
  • Fresh installs and CI never launch a daemon — both hooks no-op when no daemon was running.

Testing

• Template rendered with dummy shas and syntax-checked (ruby -c on the produced cask: Syntax OK); workflow YAML parses.
• New tests — N/A (a YAML heredoc template; the Ruby is validated by rendering it).
• Runtime behavior lands at the next release cut, when the generated cask reaches zzet/homebrew-tap; a manual brew upgrade on macOS confirms then.

Notes

• Deliberately targets the generator, not the tap file: the tap's cask is stamped out on every release and any hand edit would be overwritten.
• daemon status as the was-running probe reuses gortex's own socket resolution (no XDG path duplication in Ruby); it is bounded by the daemon's 30s control timeout.