CDK keeps waiting for the stack to deploy even when the bootstrapper failed

The way the database deployment works is :

1. Create an RDS instance
2. Create a lambda with the runtime code that creates the `pgstac` mechanism in that RDS instance
3. Create a 'custom resource', aka a one-time trigger for that lambda that happens at deployment time.

(3) happens during the deployment. However, if it fails, the CDK doesn't catch the error, but doesn't consider the deployment finished either, and keeps waiting that the stack finishes deploying until we hit the time limit.