Replace impl of Rows with ChunksExact

[RunDevelopment]

Changes:

• Make Rows a type alias for ChunksExact. Same for mut.

After #2938, Rows was only a wrapper around ChunksExact. In this PR I removed the wrapper and made it a type alias. At the cost of removing an abstraction (and having less control), I removed 150 LOC. This should also improve performance in some cases, since ChunksExact implements certain unstable traits and relies less on defaults. E.g. image.rows().nth(100) is now an efficient way to get the 100th row. Same for mut.

---

I also contemplated removing the type aliases, since they are only used once.

[RunDevelopment]

Another idea: Make Rows a view of a slice.

Basically like this:

#[derive(Copy, Clone, Debug, ...)]
pub struct Rows<'a, P> {
  pixels: &'a [P],
  width: u32,
}
impl Rows<'_, _> {
  pub fn iter(&self) -> ChunksExact;
  pub fn par_iter(&self) -> ParChunksExact;
  pub fn get(&self, y: u32) -> Option<&[P]>;
}
impl IntoIterator for Rows;
impl Index<u32> for Rows;
// similar for RowsMut

This would grow the API surface, but enable us to (1) put all row-related functionality in one place and (2) add some functionality that would be harder to justify without. E.g. a user (or we) may need mutable access to 2 different rows. With RowsMut like that, there'd be a natural way to put a get_disjoint_mut à la Vec. (This can be done by slicing pixels but is tricky.)

---

Honestly, I'm not sure what the best API for image.rows() should be. I just feel like an iterator is the wrong choice.

[197g]

Hm, providing IntoIterator without implementing Iterator itself is certainly possible? It would also justify having the struct be Copy and the indexing idea / future direction in particular for mutable rows is cute. Hm. I don't necessarily want to add breaking changes for the sake of cute, churn is a real downside. In particular here where a slice to the whole image is readily available (now) and there are considerable alternatives for representing the channel matrix such that you can slice them in both directions.