Fix clippy int_plus_one in aarch64 u8 filter bound

`while x + 16 <= w - 1` trips clippy::int_plus_one under rustc 1.95's
stricter lint. Rewrite as `x + 16 < w` — identical iteration count
(still stops once index x+16 would reach w) and avoids the `w - 1`
underflow when w == 0.

Author: enthropy7

crates/yscv-imgproc/src/ops/u8_filters.rs

@@ -1544,7 +1544,7 @@ unsafe fn sobel_row_neon(
 
     let mut x = 1usize;
 
-    while x + 16 <= w - 1 {
+    while x + 16 < w {
         // Load 16 u8 at offsets x-1, x, x+1 for each row
         let t_l = vld1q_u8(row0.add(x - 1));
         let t_c = vld1q_u8(row0.add(x));