[BUG] ParallelCluster database stack should include slurmdbd instance and slurmdbd

[cartalla]

Describe the bug
Currently ParallelCluster provides a CloudFormation template to create the Slurm database.
This stack only creates an RDS database, but doesn't create an instance to run the slurmdbd daemon.
This doesn't follow the recommended architecture and limits the ability of multiple cluster to share the same accounting database.
Currently ParallelCluster runs a slurmdbd daemon on the controller of each cluster that is using the database.
This isn't the recommended architecture and I'm not sure if this will work.