fix: prevent set -e from killing entrypoint on DB detection

The python3 DB detection script exits with code 1 (empty) or 2 (legacy),
but entrypoint has set -e which kills the script on any non-zero exit.
Fixed by using '|| _db_state=$?' pattern to safely capture exit code
without triggering set -e.

docker/backend.entrypoint.sh

@@ -152,18 +152,19 @@ install_certificates
 install_playwright
 
 # Check DB state: empty, legacy (tables exist but no alembic), or managed
+# || prevents set -e from killing the script on non-zero exit from detection script
+_db_state=0
 python3 -c "
 import os
 from sqlalchemy import create_engine, inspect
 e = create_engine(os.environ['DATABASE_URL'])
 insp = inspect(e)
 if 'alembic_version' in insp.get_table_names():
-    exit(0)   # managed → normal upgrade
+    exit(0)
 if not insp.get_table_names():
-    exit(1)   # empty → fresh upgrade
-exit(2)       # legacy → stamp head
-" 2>/dev/null
-_db_state=$?
+    exit(1)
+exit(2)
+" 2>&1 || _db_state=$?
 
 if [ $_db_state -eq 0 ]; then
   echo "[entrypoint] Running Alembic migrations (incremental)..."