planner: keep unfoldable exprs when pruning columns for ORDER BY items

[lysu]

What problem does this PR solve?

fixes #9033

What is changed and how it works?

pruning column logic will do prune if order by expression with zero columns(it's right for sql like order by abs(1)), but rand() shouldn't be pruned, so do skip logic.

Check List

Tests

• Unit test

Code changes

• implement change

Side effects

• N/A

Related changes

• Need to cherry-pick to the release branch

---

This change is 

[eurekaka]

LGTM

[lysu]

/run-all-tests

[lysu]

@qw4990 @winoros have changed to use ConstItem to determine prune PTAL thx~

[lysu]

we need:
desc select * from t1 where t1.a in (select t2.a as a from t2 where t2.b > t1.b order by t1.b limit 1);
id	count	task	operator info
Apply_15	9990.00	root	semi join, inner:Selection_19, equal:[eq(test.t1.a, test.t2.a)]
├─TableReader_18	9990.00	root	data:Selection_17
│ └─Selection_17	9990.00	cop	not(isnull(test.t1.a))
│   └─TableScan_16	10000.00	cop	table:t1, range:[-inf,+inf], keep order:false, stats:pseudo
└─Selection_19	0.80	root	not(isnull(test.t2.a))
  └─Limit_20	1.00	root	offset:0, count:1
    └─TableReader_26	1.00	root	data:Limit_25
      └─Limit_25	1.00	cop	offset:0, count:1
        └─Selection_24	1.00	cop	gt(test.t2.b, test.t1.b)
          └─TableScan_23	1.25	cop	table:t2, range:[-inf,+inf], keep order:false, stats:pseudo
desc select * from t1 where t1.a i
but got:
desc select * from t1 where t1.a in (select t2.a as a from t2 where t2.b > t1.b order by t1.b limit 1);
id	count	task	operator info
Apply_14	9990.00	root	semi join, inner:Selection_18, equal:[eq(test.t1.a, test.t2.a)]
├─TableReader_17	9990.00	root	data:Selection_16
│ └─Selection_16	9990.00	cop	not(isnull(test.t1.a))
│   └─TableScan_15	10000.00	cop	table:t1, range:[-inf,+inf], keep order:false, stats:pseudo
└─Selection_18	0.80	root	not(isnull(test.t2.a))
  └─TopN_19	1.00	root	test.t1.b:asc, offset:0, count:1
    └─TableReader_25	1.00	root	data:TopN_24
      └─TopN_24	1.00	cop	test.t1.b:asc, offset:0, count:1
        └─Selection_23	8000.00	cop	gt(test.t2.b, test.t1.b)
          └─TableScan_22	10000.00	cop	table:t2, range:[-inf,+inf], keep order:false, stats:pseudo

it seems there are another order by pruning bug when subquery order by correlate column - - but solved by this PR

[lysu]

/run-all-tests

[qw4990]

LGTM

[lysu]

/run-integration-ddl-test
/run-sqllogic-test-2

[zz-jason]

we need:
desc select * from t1 where t1.a in (select t2.a as a from t2 where t2.b > t1.b order by t1.b limit 1);
id	count	task	operator info
Apply_15	9990.00	root	semi join, inner:Selection_19, equal:[eq(test.t1.a, test.t2.a)]
├─TableReader_18	9990.00	root	data:Selection_17
│ └─Selection_17	9990.00	cop	not(isnull(test.t1.a))
│   └─TableScan_16	10000.00	cop	table:t1, range:[-inf,+inf], keep order:false, stats:pseudo
└─Selection_19	0.80	root	not(isnull(test.t2.a))
  └─Limit_20	1.00	root	offset:0, count:1
    └─TableReader_26	1.00	root	data:Limit_25
      └─Limit_25	1.00	cop	offset:0, count:1
        └─Selection_24	1.00	cop	gt(test.t2.b, test.t1.b)
          └─TableScan_23	1.25	cop	table:t2, range:[-inf,+inf], keep order:false, stats:pseudo
desc select * from t1 where t1.a i
but got:
desc select * from t1 where t1.a in (select t2.a as a from t2 where t2.b > t1.b order by t1.b limit 1);
id	count	task	operator info
Apply_14	9990.00	root	semi join, inner:Selection_18, equal:[eq(test.t1.a, test.t2.a)]
├─TableReader_17	9990.00	root	data:Selection_16
│ └─Selection_16	9990.00	cop	not(isnull(test.t1.a))
│   └─TableScan_15	10000.00	cop	table:t1, range:[-inf,+inf], keep order:false, stats:pseudo
└─Selection_18	0.80	root	not(isnull(test.t2.a))
  └─TopN_19	1.00	root	test.t1.b:asc, offset:0, count:1
    └─TableReader_25	1.00	root	data:TopN_24
      └─TopN_24	1.00	cop	test.t1.b:asc, offset:0, count:1
        └─Selection_23	8000.00	cop	gt(test.t2.b, test.t1.b)
          └─TableScan_22	10000.00	cop	table:t2, range:[-inf,+inf], keep order:false, stats:pseudo

it seems there are another order by pruning bug when subquery order by correlate column - - but solved by this PR

This change is because CorrelatedColumn.ConstItems() returns false, which is incorrect, it's should be true.

[zz-jason]

LGTM