Add keras-tuner like interfaces for Hyperparam optimization

Keras-tuner (https://github.com/keras-team/keras-tuner) is a hyperparameter tuner for `tf.keras`. We like to take insights from this framework. In particular, we like to design the following interface in our framework

1) Develop and expose a Hyperparam class like `TPE`, `Hyperband` etc to the user. The user supplies the `objective_func` while creating this object something like

```
tuner = TPE(
objective_func,
objective='val_accuracy',
max_trials=5,
executions_per_trial=3,)
```

2) The user invokes hyperparameter search by calling `search()`

```
tuner.search()
```

3) The user can get results through the following utility functions

```
models = tuner.get_best_models(num_models=2)
```

or

```
tuner.results_summary()
```

In addition, we may like to provide pre-made tunable applications like they do in keras-tuner.

Start by reviewing the proposed TPE and Hyperband interfaces and comparing them with keras-tuner and tf.keras. Define the scope for objective_func, search(), get_best_models(), and results_summary(), then identify the existing framework entry points where these APIs would fit; the work is done when the agreed interfaces and any tunable applications are implemented and usable.

Written by the indexing model from the issue text.