Built-in utilities for parameter fitting

We already have examples/recipes, but it could be even easier for people if there were built-in utilities:

```python
@memo
def f(alpha, beta): ...

f.grid_search(
  loss=lambda pred: …,
  grid=(np.linspace(…), np.linspace(…), …)
)

f.gradient_fit(
  loss=lambda pred: …,
  init=(0, 0.5),
  step_size=…,
  num_iterations=…
)
```