Fix RDS parameter group upgrading to create_before_destroy

When upgrading to v0.10.0, we hit an undocumented breaking change - the
change of the RDS parameter group name necessitates Terraform recreate
the resource. But, the parameter group cannot be destroyed while the
running database is currently using it.

By creating a new parameter group first (before destroying the old one),
we can handle upgrades to the parameter group without causing breaking
changes.

infra/modules/database/main.tf

@@ -94,4 +94,8 @@ resource "aws_rds_cluster_parameter_group" "rds_query_logging" {
     # Logs all statements that run 100ms or longer
     value = "100"
   }
+
+  lifecycle {
+    create_before_destroy = true
+  }
 }