libnm: use nm_ascii_is_ctrl_or_del() in nm_utils_fixup_product_string() to preserve UTF-8

On architectures where "char" is signed, the check "ch < ' '" is also
TRUE for characters greater than 127 (that is, UTF-8 characters).

Let's preserve valid UTF-8 characters and don't clear them.

Also note that already before we filtered out invalid UTF-8 sequences,
so if we encounter here a character > 127, it is part of a valid UTF-8
sequence.

src/libnm-client-impl/nm-libnm-utils.c

@@ -166,7 +166,7 @@ _fixup_string(const char *       desc,
             in_paren = TRUE;
         else if (*p == ')')
             in_paren = FALSE;
-        else if (NM_IN_SET(*p, '_', ',') || *p < ' ' || in_paren) {
+        else if (NM_IN_SET(*p, '_', ',') || nm_ascii_is_ctrl_or_del(*p) || in_paren) {
             /* pass */
         } else
             continue;