Add method to modify elements in place

[bluss]

A value-to-value mapper would be convenient, something like array.apply(f32::abs) it could even be called lift.

[bluss]

We want to find a nice convention to make it consistent.

Maybe something like this, saying map for an operation that makes a new array, and apply for modifying in place. Since all the float functions are by value "|A| -> A", it would be important to offer this.

 .map(&self, |&A| -> B) -> OwnedArray<B, D>
 .mapv(&self, |A| -> B) -> OwnedArray<B, D>
 .apply(&mut self, |&mut A|)
 .applyv(&mut self, |A| -> A)
 .visit(&self, |&A|)

[jdroenner]

Maybe the map operations could be lazy like iterator adapters. A MapView(or something like this) would have a reference/view of an array and a map function. This would allow to use get(dim)without processing the total array first (could be useful for sampling). Also chaining of multiple map operations could be possible. A function like collect could be used to materialize into a new OwnedArray. This could also be interesting for #120 or #127.

[bluss]

That's a nice idea. It's similar to this #88 (comment)

Making an operation an object that itself can participate in operations on arrays makes particular sense in an even more generic-heavy implementation. For that reason ndarray will not take that route now, I think.