fixing k8s node types for trn1 instance types

[clumsy]

Now that we use K8S_ITYPE for multi-node job's instance_type for AWS resources, it has to match the official label or we get the following error with aws_batch_scheduler:

botocore.errorfactory.ClientException: An error occurred (ClientException) when calling the RegisterJobDefinition operation: Error executing request, Exception : Instance type can only be one of [m6g.xlarge, ..., c7g.large, trn1.2xlarge, c7g.12xlarge, ..., trn1.32xlarge, ..., g4dn.4xlarge, trn1n.32xlarge, ..., r7g.medium].

Intentionally not keeping an alias to previous labels, e.g. aws_trn1.32xl because torchx suggests the new one.

Test plan:
Fixed unit tests

[clumsy]

This is a regression, I'm afraid @kiukchung

[kiukchung]

LGTM thanks