MDEV-23350 ROUND(bigint_22_or_longer) returns a wrong data type

The condition in Item_func_round::fix_arg_int() to decide whether:
- we can preserve the data type of args[0] versus
- the result can go outside of the args[0] data type
was wrong.

The data type of the first argument can be preserved in these cases:
- TRUNCATE(x, n)
- ROUND(x, n>=0)

Fixing the condition accordingly.

mysql-test/main/type_int.test

@@ -295,6 +295,7 @@ DROP TABLE t1;
 
 --echo #
 --echo # MDEV-23337 Rounding functions create a wrong data type for numeric input
+--echo # MDEV-23350 ROUND(bigint_22_or_longer) returns a wrong data type
 --echo #
 
 DELIMITER $$;
@@ -304,7 +305,7 @@ BEGIN
   EXECUTE IMMEDIATE REPLACE('CREATE TABLE t1 (a TYPE)', 'TYPE', t);
   INSERT IGNORE INTO t1 VALUES (-100000000000000000000000000000);
   INSERT IGNORE INTO t1 VALUES (100000000000000000000000000000);
-  CREATE  TABLE t2 AS SELECT a, ROUND(a), TRUNCATE(a,0), FLOOR(a), CEILING(a) FROM t1;
+  CREATE  TABLE t2 AS SELECT a, ROUND(a), ROUND(a,1), ROUND(a,10), TRUNCATE(a,0), FLOOR(a), CEILING(a) FROM t1;
   SHOW CREATE TABLE t2;
   SELECT * FROM t2;
   DROP TABLE t1, t2;
@@ -318,12 +319,22 @@ CALL p1('smallint');
 CALL p1('mediumint');
 CALL p1('int');
 CALL p1('bigint');
+CALL p1('bigint(20)');
+CALL p1('bigint(21)');
+CALL p1('bigint(22)');
+CALL p1('bigint(23)');
+CALL p1('bigint(30)');
 
 CALL p1('tinyint unsigned');
 CALL p1('smallint unsigned');
 CALL p1('mediumint unsigned');
 CALL p1('int unsigned');
 CALL p1('bigint unsigned');
+CALL p1('bigint(20) unsigned');
+CALL p1('bigint(21) unsigned');
+CALL p1('bigint(22) unsigned');
+CALL p1('bigint(23) unsigned');
+CALL p1('bigint(30) unsigned');
 --horizontal_results
 
 DROP PROCEDURE p1;

sql/item_func.cc

@@ -2472,7 +2472,8 @@ void Item_func_round::fix_arg_int(const Type_handler *preferred,
     Longlong_hybrid val1= args[1]->to_longlong_hybrid();
     if (args[1]->null_value)
       fix_length_and_dec_double(NOT_FIXED_DEC);
-    else if ((!val1.to_uint(DECIMAL_MAX_SCALE) && truncate) ||
+    else if (truncate ||
+             !val1.neg() /* ROUND(x, n>=0) */ ||
              args[0]->decimal_precision() < DECIMAL_LONGLONG_DIGITS)
     {
       // Here we can keep INT_RESULT