Deployment: Fix npm command issue #300

Describe the bug
The v0.6.1 production release failed again in "Wait for SSM command to finish", this time with:

.../_script.sh: line 1: npm: command not found
failed to run commands: exit status 127

Root cause

• fix(workflows): Fix production release workflow #299 fixed the earlier $HOME not set error (Deployment: Fix $HOME issue #298) by setting HOME=/root and keeping the deploy running as root.
• With that, git fetch and git checkout succeed, but root has no npm on its PATH. On the production host, node is installed through NVM under the deploy user (/home/<user>/.nvm), not system-wide.
• The app also runs in the deploy user's pm2. So even with npm found, a root pm2 restart would not find the app.
• Before chore(deployment): Improve build process #292, production loaded NVM from the deploy user's home and ran sudo -iu <user> pm2 restart. deploy-staging.yml still does this and deploys fine.

Impact

• Production is still serving v0.5.0. v0.6.0 and v0.6.1 are not live.
• On the server, the repo is now checked out at 7c29ec6 (v0.6.1), but .next is still the old build and pm2 was not restarted.

To Reproduce

1. Push a release tag (e.g. v0.6.1).
2. "Deploy Kaapi to EC2 Production" fails at "Wait for SSM command to finish" with npm: command not found.

Expected behavior
The release tag deploys to production: checkout, npm ci, build, and pm2 restart all succeed.

Fix
Use the same remote command as deploy-staging.yml. Set ROOT_USER from the existing USER secret in the AWS_ENV environment, then export HOME=/home/$ROOT_USER, load NVM from that user's home, and run sudo -iu $ROOT_USER pm2 restart. After merging, cut a new release (v0.6.2). A tag runs the workflow file from its own commit, so re-running v0.6.1 won't pick up the fix.