feat: add Hetzner-hosted todo application stack

Implement the server-rendered Go todo application with PostgreSQL persistence, Tailwind styling, health checks, and test coverage.

Provision production and staging infrastructure with OpenTofu, including private networking, static IPs, Object Storage backups, and load balancing.

Configure Rocky Linux hosts with Ansible, rootless Podman, PostgreSQL, WAL-G backup automation, staging restores, and generated inventory.

ansible/roles/wal_g/templates/db-sync.sh.j2

@@ -0,0 +1,54 @@
+#!/bin/bash
+set -Eeuo pipefail
+source /etc/wal-g/db-sync.env
+
+expected_pgdata="/var/lib/pgsql/data"
+if [[ "$PGDATA" != "$expected_pgdata" || ! -d "$PGDATA" || "$(stat -c %U "$PGDATA")" != "postgres" ]]; then
+  echo "Refusing destructive restore for unexpected PGDATA: $PGDATA" >&2
+  exit 1
+fi
+if [[ "$WALG_S3_PREFIX" != s3://*/prod ]]; then
+  echo "Refusing restore from a non-production prefix" >&2
+  exit 1
+fi
+
+systemctl --user --machine="${APP_USER}@" stop app-todo.service || true
+systemctl stop "$POSTGRES_SERVICE"
+
+shopt -s dotglob nullglob
+contents=("$PGDATA"/*)
+if (( ${#contents[@]} > 0 )); then
+  rm -rf --one-file-system -- "${contents[@]}"
+fi
+
+sudo -u postgres --preserve-env=AWS_ACCESS_KEY_ID,AWS_SECRET_ACCESS_KEY,AWS_ENDPOINT,AWS_REGION,AWS_S3_FORCE_PATH_STYLE,WALG_S3_PREFIX,WALG_COMPRESSION_METHOD \
+  /usr/local/bin/wal-g backup-fetch "$PGDATA" LATEST
+chown -R postgres:postgres "$PGDATA"
+chmod 0700 "$PGDATA"
+install -o postgres -g postgres -m 0600 /etc/todo-postgresql/postgresql.conf "$PGDATA/postgresql.conf"
+install -o postgres -g postgres -m 0600 /etc/todo-postgresql/pg_hba.conf "$PGDATA/pg_hba.conf"
+
+cat >"$PGDATA/postgresql.auto.conf" <<'RECOVERY'
+restore_command = '/usr/local/sbin/wal-g-postgres wal-fetch %f %p'
+recovery_target_timeline = 'latest'
+RECOVERY
+touch "$PGDATA/recovery.signal"
+chown postgres:postgres "$PGDATA/postgresql.auto.conf" "$PGDATA/recovery.signal"
+
+systemctl start "$POSTGRES_SERVICE"
+for _ in {1..120}; do
+  if sudo -u postgres psql --dbname=postgres --tuples-only --no-align --command="SELECT NOT pg_is_in_recovery()" 2>/dev/null | grep -qx t; then
+    break
+  fi
+  sleep 2
+done
+if ! sudo -u postgres psql --dbname=postgres --tuples-only --no-align --command="SELECT NOT pg_is_in_recovery()" | grep -qx t; then
+  echo "PostgreSQL did not complete recovery" >&2
+  exit 1
+fi
+
+sudo -u postgres psql --dbname=postgres --set="role=$APP_DB_USER" --set="password=$APP_DB_PASSWORD" <<'SQL'
+SELECT format('ALTER ROLE %I LOGIN PASSWORD %L', :'role', :'password') \gexec
+SQL
+
+systemctl --user --machine="${APP_USER}@" start app-todo.service