needless_collect triggers inside for loop, can be questionable

    let v = vec![1, 2, 3, 4].into_iter().map(|x| /*very expensive closure */ x).collect::<Vec<_>>();
   
    for i in &[3, 4] {
        if v.contains(&i) {
            println!("got {}", i);
        }
    }

Right now, we are running the expensive closure only N (vec.len() times.
If we move it into into the for-loop, this could easily become N (vec.len() * M (for loop) times

fn main() {
    for i in &[3, 4] {
        if vec![1, 2, 3, 4].into_iter().map(|x| /* very expensive */ x).any(|x| x == *i) {
            println!("got {}", i);
        }
    }
}

it might be cheaper to just run the closure once up front for all the elements and "cache" the results in the vec.

Start by locating the needless_collect lint and its existing tests, then compare them with the two Rust examples in this issue. Determine the intended behavior when the collected value is searched repeatedly inside a for loop, and add regression coverage that makes the accepted outcome explicit.