bugfix: Fix "default" ecs cluster auto-create logic

[sparrc]

Summary

This fixes agent's ClusterNotFound error handling, which is returned as a ClientException in the case that the agent calls RCI and the cluster does not exist.

Testing

A new unit test file was added for testing the entire errors package, including this function.

Manual test was run to confirm that an account/region with no default cluster now has one auto-created and agent is able to register with the default cluster:

level=error time=2025-06-19T23:46:15Z msg="Unable to register as a container instance with ECS" error="operation error ECS: RegisterContainerInstance, https response error StatusCode: 400, RequestID: NNN, ClientException: Cluster not found."
level=error time=2025-06-19T23:46:15Z msg="Received terminal error from RegisterContainerInstance call, exiting" error="operation error ECS: RegisterContainerInstance, https response error StatusCode: 400, RequestID: NNN, ClientException: Cluster not found."
level=info time=2025-06-19T23:46:15Z msg="Successfully created a cluster" cluster="default"

New tests cover the changes: yes

Description for the changelog

bugfix: Fix "default" ecs cluster auto-create logic

Licensing

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[sparrc]

This was previously part of this function in aws sdk v1, see previous commit

amazon-ecs-agent/agent/vendor/github.com/aws/amazon-ecs-agent/ecs-agent/api/errors/errors.go

Line 65 in a19994a

func IsClusterNotFoundError(err error) bool {

[sparrc]

Unrelated to this PR, but this unit test fails on instances with more than 64GB of memory, because the mem value returned from getHostMemoryInMiB overflows uint16 used on line 789 below.