feat: add eval progress tracking

[joshuatowner]

Add eval progress tracking

Replaces the Rich Live spinner in the terminal wait() path with plain-text status prints that stream correctly to agents and piped processes.

Problem

The eval wait() terminal path used a Rich Live spinner that:

• Outputs ANSI escape sequences agents can't parse
• Uses transient=True so output disappears after completion
• Only shows "Current status: Executing" with no step details

Changes

• Remove Rich Live spinner from terminal wait() path
• Print pipeline name and execution ARN at start
• Print step transitions with ✓/⋯, durations, and job ARNs on each poll
• Print "Running... Xs" elapsed timer for in-progress steps
• Print S3 output path on success
• Print failed step details with log group and CloudWatch link on failure
• All prints use flush=True for agent compatibility
• Log group derived from job ARN type (training/processing/transform)

Jupyter path is unchanged.

Example output

Evaluation started: eval-meta-5190c33c
Pipeline: SagemakerEvaluation-LLMAJEvaluation-565d3249
Execution ARN: arn:aws:sagemaker:us-west-2:099324990371:pipeline/.../execution/rstzamudixmz

--------------------------------------

Status Transitions:
  ✓ CreateEvaluationAction: Succeeded (3.0s)
  ⋯ EvaluateCustomInferenceModel: Executing (Running... 17s)
    Job ARN: arn:aws:sagemaker:us-west-2:099324990371:training-job/CustomInference-rstzamudixmz-6ZP4YbszZc

Status: Executing (Elapsed: 21.5s)

On failure:

Failed
Failure reason: Step failure: One or multiple steps failed.

Failed step: EvaluateCustomInferenceModel
Failure reason: ClientError: No S3 objects found under S3 URL...
Job ARN: arn:aws:sagemaker:us-west-2:099324990371:training-job/CustomInference-xyz
Log group: /aws/sagemaker/TrainingJobs
Log stream prefix: CustomInference-xyz
CloudWatch Logs: https://us-west-2.console.aws.amazon.com/cloudwatch/...

Testing

• 382 existing unit tests pass (0 regressions)
• 4 new tests covering start, step transitions, success, and failure output
• Manually validated with real LLM-as-Judge eval jobs