fix(query): fold constant subquery to build filter plan instead of join plan

[b41sh]

I hereby agree to the terms of the CLA available at: https://docs.databend.com/dev/policies/cla/

Summary

Fold a subquery that returns a constant value to a constant value scalar, convert the join plan to a filter plan, which can be filtered using indexes to avoid reading the full amount of data and speed up the query.

for example:
main

root@0.0.0.0:48000/default> explain select * from numbers(10) as t where t.number in (select * from unnest([1,2]));
-[ EXPLAIN ]-----------------------------------
HashJoin
├── output columns: [t.number (#0)]
├── join type: LEFT SEMI
├── build keys: [CAST(subquery_2 (#2) AS UInt64 NULL)]
├── probe keys: [CAST(t.number (#0) AS UInt64 NULL)]
├── filters: []
├── estimated rows: 10.00
├── EvalScalar(Build)
│   ├── output columns: [unnest([1, 2]) (#2)]
│   ├── expressions: [get(1)(unnest([1, 2]) (#1))]
│   ├── estimated rows: 3.00
│   └── ProjectSet
│       ├── output columns: [unnest([1, 2]) (#1)]
│       ├── estimated rows: 3.00
│       ├── set returning functions: unnest([1, 2])
│       └── DummyTableScan
└── TableScan(Probe)
    ├── table: default.system.numbers
    ├── output columns: [number (#0)]
    ├── read rows: 10
    ├── read size: < 1 KiB
    ├── partitions total: 1
    ├── partitions scanned: 1
    ├── push downs: [filters: [], limit: NONE]
    └── estimated rows: 10.00

25 rows explain in 0.005 sec. Processed 0 rows, 0 B (0 row/s, 0 B/s)

this PR

root@0.0.0.0:48000/default> explain select * from numbers(10) as t where t.number in (select * from unnest([1,2]));
-[ EXPLAIN ]-----------------------------------
Filter
├── output columns: [t.number (#0)]
├── filters: [(t.number (#0) = 1 OR t.number (#0) = 2)]
├── estimated rows: 0.01
└── TableScan
    ├── table: default.system.numbers
    ├── output columns: [number (#0)]
    ├── read rows: 10
    ├── read size: < 1 KiB
    ├── partitions total: 1
    ├── partitions scanned: 1
    ├── push downs: [filters: [(numbers.number (#0) = 1 OR numbers.number (#0) = 2)], limit: NONE]
    └── estimated rows: 10.00

13 rows explain in 0.007 sec. Processed 0 rows, 0 B (0 row/s, 0 B/s)

• part of: bug: Full Table Scan Issue with UNNEST(SPLIT(...)) and IN/CTE Queries in Databend #17440

Tests

• Unit Test
• Logic Test
• Benchmark Test
• No Test - Explain why

Type of change

• Bug Fix (non-breaking change which fixes an issue)
• New Feature (non-breaking change which adds functionality)
• Breaking Change (fix or feature that could cause existing functionality not to work as expected)
• Documentation Update
• Refactoring
• Performance Improvement
• Other (please describe):

---

This change is