numfmt: reject an oversized --from-unit/--to-unit instead of panicking

parse_unit_size computed the unit size as `n * multiplier` with an
unchecked multiplication. A large numeric part with a suffix overflows
usize: when the product wraps to exactly 0 (e.g. `--to-unit=18014398509481984Ki`,
2^54 * 1024 = 2^64), the zero unit later reaches `integer % to_unit` and
aborts with a remainder-by-zero; a merely-huge value overflows the
multiply itself under overflow-checks.

Use `checked_mul` so an out-of-range product falls through to the
existing "invalid unit size" error (exit 1), matching GNU numfmt, whose
`unit_to_umax` rejects a value that overflows uintmax_t.

Author: leeewee

src/uu/numfmt/src/numfmt.rs

@@ -200,8 +200,12 @@ fn parse_unit_size(s: &str) -> Result<usize> {
             if number.is_empty() {
                 return Ok(multiplier);
             }
-            if let Ok(n) = number.parse::<usize>() {
-                return Ok(n * multiplier);
+            if let Some(size) = number
+                .parse::<usize>()
+                .ok()
+                .and_then(|n| n.checked_mul(multiplier))
+            {
+                return Ok(size);
             }
         }
     }

tests/by-util/test_numfmt.rs

@@ -878,7 +878,13 @@ fn test_to_unit() {
 #[test]
 fn test_invalid_unit_size() {
     let commands = vec!["from", "to"];
-    let invalid_sizes = vec!["A", "0", "18446744073709551616"];
+    let invalid_sizes = vec![
+        "A",
+        "0",
+        "18446744073709551616",
+        "18446744073709551615K",
+        "18014398509481984Ki",
+    ];
 
     for command in commands {
         for invalid_size in &invalid_sizes {