Remove the num-iter dependency.

[RazrFalcon]

num-iter is used just for range_step iterator. By using std::iter::step_by we can remove a lot of bloat:

`-- num-iter v0.1.39
    |-- num-integer v0.1.41
    |   `-- num-traits v0.2.8
    |       [build-dependencies]
    |       `-- autocfg v0.1.4
    |   [build-dependencies]
    |   `-- autocfg v0.1.4 (*)
    `-- num-traits v0.2.8 (*)
    [build-dependencies]
    `-- autocfg v0.1.4 (*)

This also decreases the build time: 4.9s vs 6.7s

[HeroicKatora]

Do we even need this anymore? (start..stop).step_by(step) seems to achive the very same thing from the standard library. Is that a performance difference?

[RazrFalcon]

I've wanted to use step_by too, but it can step only by usize and this code uses isize.

[HeroicKatora]

bit_depth (the step argument) is a u8 initially, so that should be no problem. In fact, the line should have bit_depth.into() instead of bit_depth as isize to avoid making it seems like a lossy conversion.

[RazrFalcon]

range_step used more than once.

[HeroicKatora]

Oh, right 👍

[HeroicKatora]

@RazrFalcon I'm kind of confused now. You said it is required to be used for isize as well but all usages cast to usize eventually. It seems all using functions expect the actual range of values to always be positive because they are used as indices (only within utils.rs). So, it should be possible to replace the isize in all iterators with carful application of .rev(), .step_by() and RangeInclusive (0..=buf.len()).

[HeroicKatora]

E.g.

pub fn expand_trns_line(buf: &mut[u8], trns: &[u8], channels: usize) {
    let i = (0..=buf.len()/(channels+1)*channels - channels).rev().step_by(channels);
    let j = (0..=buf.len() - (channels+1)).rev().step_by(channels+1);
    let channels = channels as usize;
    for (i, j) in i.zip(j) {
        if &buf[i..i+channels] == trns {
            buf[j+channels] = 0
        } else {
            buf[j+channels] = 0xFF
        }
        for k in (0..channels).rev() {
            buf[j+k] = buf[i+k];
        }
    }
}

[RazrFalcon]

Yes. I've simply replaced the current functions without investigating the code that much. I'm not familiar with the code, so I don't know why isize is used. I guess it simply steps from n to 0.

I will try to change it to step_by.

[RazrFalcon]

Done.

[HeroicKatora]

Great work, thank you! Want to join image-rs? You can be invited to the image-rs collaborators team, which allows you access to this and the other repos if you want. See the contribution policy and code of conduct. (I'll guess you have a ton of work from resvg already but asking won't hurt).

[RazrFalcon]

Thanks, but I'm pretty busy with my own project already 😄

[HeroicKatora]

If you ever change your mind, just ask (and probably link to this PR). Happy coding 😄