Special-case the handling of DB Cluster- and DB Instance-level metrics in RDS

[ripta]

Specifically, with a configuration like such:

[[exportconfigs]]

   namespace = "AWS/RDS"
   name = "CPUUtilization"
   dimensionsMatch   = { DBInstanceIdentifier = "..."    }
   dimensions = ["DBInstanceIdentifier"]
   statistics = ["Average"]

[[exportconfigs]]

   namespace = "AWS/RDS"
   name = "CPUUtilization"
   dimensionsMatch   = { DBClusterIdentifier = "..."    }
   dimensions = ["DBClusterIdentifier" , "Role"]
   statistics = ["Average"]

the previous behavior would have generated the following error:

Namespace=AWS/RDS Name=CPUUtilization: a previously registered descriptor with the same fully-qualified name as Desc{fqName: "aws_rds_utilization_average", help: "", constLabels: {}, variableLabels: [cluster_identifier role]} has different label names or a different help string

The patch would allow the configuration above, with the caveat that metrics are now renamed from, e.g., aws_rds_utilization_average{instance_identifier="shared"} to aws_rds_utilization_instance_average{instance_identifier="shared"}. Similarly, a cluster would have _cluster instead of _instance.