Create an `mpy.optimizer` package with an `mpy.optimizer.RandomOptimizer`

```python
exp = mpy.Float(0, 10)
opt = mpy.optimizer.RandomOptimizer(exp,sampler=None, rng=42)
for config, frozen_expression in opt.sample(size=10):
  pytorch_model = frozen_expression.evaluate()
  val_acc = train(pytorch_model)
  opt.tell(config, val_acc)
  # it means the optimizer as and ask/tell interface 
  # and the ask is used in sample to iterate
```