Provide metrics for non-zero exit codes

This is more of an RFC, than a bug report. Please let me know if there is another mechanism that I should be using to submit a feature request :slightly_smiling_face:

### Summary

Currently, there is no easy way to monitor how many containers are exiting with non-zero [exit codes](https://github.com/aws/amazon-ecs-agent/blob/master/agent/sighandlers/exitcodes/exitcodes.go). We should provide a mechanism to (optionally) push the exit status code as a count metric to CloudWatch.

### Description

It is probably best if this is enabled via a [configuration option](https://docs.aws.amazon.com/AmazonECS/latest/developerguide/ecs-agent-config.html) in `/etc/ecs/ecs.config`, as cloudwatch metrics can be unnecessarily costly if you are not making use of them.
For me, it would be enough to just have a `ExitError` and `ExitSuccess` count, but others may prefer having separate metrics for each possible [exit code](https://github.com/aws/amazon-ecs-agent/blob/master/agent/sighandlers/exitcodes/exitcodes.go).

### Expected Behavior
Push metrics to CloudWatch.

### Observed Behavior
N/A

### Environment Details

Any [ECS-optimised AMI](https://docs.aws.amazon.com/AmazonECS/latest/developerguide/ecs-optimized_AMI.html)

### Supporting Log Snippets

N/A

Start with agent/sighandlers/exitcodes/exitcodes.go and the ECS agent configuration documentation for /etc/ecs/ecs.config. Review how exit statuses are observed and how CloudWatch metrics are currently configured, then determine the metric shape and configuration behavior needed for optional exit-status counts. Done means the chosen metrics can be enabled and pushed to CloudWatch as described.

Written by the indexing model from the issue text.