Improve Zig Zag conversion

[ChocolateLoverRaj]

Single allocation, no reallocations

This takes advantage of each character always being 1 byte, which is guaranteed by the LeetCode constraints.

Instead of actually constructing a zigazag this code just finds the Math pattern behind every row to construct the final output.

[jusexton]

Hey @ChocolateLoverRaj! Thanks for the allocation improvement. Once the linting issues are fixed it would also be nice to reuse the logic for the top and bottom rows. Something like this:

fn append(o: &mut String, bytes: &[u8], start: usize, step: usize) {
    let mut i = 0;
    while let Some(&c) = bytes.get(start + i * step) {
        o.push(c as char);
        i += 1;
    }
}

let num_rows = num_rows as usize;
let bytes = s.as_bytes();
let step = match num_rows {
    1 => 1,
    n => n * 2 - 2,
};
let mut o = String::with_capacity(s.len());

// Top row
append(&mut o, bytes, 0, step);

// Middle row

// Bottom row
if num_rows > 1 {
    append(&mut o, bytes, num_rows - 1, step);
}