AWS::ElastiCache::ReplicationGroup - failed node type update is recorded as success by CloudFormation

### Name of the resource

AWS::ElastiCache::ReplicationGroup

### Resource Name

_No response_

### Issue Description

As explained in https://repost.aws/knowledge-center/elasticache-redis-change-node-errors, if, for example, you apply a change to reduce the instance size to one with insufficient memory, ElastiCache will fail to deploy this change, and you get a message in the "Events" section of ElastiCache to say:

> Failed to scale down to cache node type Replication Group [myreplicationgroup] because the node has insufficient memory. Please select a different node type or reduce current memory usage and retry

However, CloudFormation doesn't detect this as an error and records the resource change as "UPDATE_COMPLETE" even though the change has not been applied and there is now drift between CloudFormation and ElastiCache.

### Expected Behavior

If the update fails to apply, CloudFormation should notice this and record that the update failed.

### Observed Behavior

The update fails to apply but CloudFormation says it completed.

### Test Cases

- Create a ReplicationGroup using a large instance type
- Use it such that it consumes lots of memory
- Apply a change using CloudFormation to use a small instance type
- See that CloudFormation sees the change as successful, but that the change isn't actually applied, and there's an error in the "Events" tab in ElastiCache.

### Other Details

_No response_