KEP-3562: [OptimizationJob] Add support for Early Stopping (Pruning), Hyperband, and PBT

### What you would like to be added?

Context:
In [KEP-3562](https://github.com/kubeflow/trainer/pull/3565), Phase 1 focuses on standard search strategies (Random, Grid).

Problem:
Running underperforming trials to full completion wastes compute resources and increases cloud costs.

Future Goals:

- Add `PruneAlgorithms` to the `OptimizationJob` schema (e.g., Hyperband, Population Based Training).

- Implement intermediate trial metric reporting from TrainJob workers back to the controller/Optuna engine.

- Add controller logic to terminate unpromising TrainJobs early.

Relates to KEP-3562 Master Tracking [Issue](https://github.com/kubeflow/trainer/issues/3562).

### Why is this needed?

Early stopping and advanced pruning algorithms save up to 80% of compute resources during hyperparameter searches by stopping poor trials early.

### Love this feature?

Give it a 👍 We prioritize the features with most 👍