Address failures that break telemetry

[rtpascual]

Describe the bug

In the scenario when failures happen before the CLI gets to send telemetry, we are unable to figure out impact.

A suggestion may be to dynamically import yargs and lift it up in order to have these types of failures go through our error handling.

See https://github.com/aws-amplify/amplify-backend/actions/runs/10876578911/job/30176690831#step:6:1133 for an example of an error that happened when loading CLI before it runs a command.