refactor: use binary for non-utf8 string type

[andylokandy]

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

Summary

1. Aggregate state: String -> Binary
2. Bloom index state: String -> Binary
3. Group by hash: String -> Binary
4. Column to arrow: String: LargeBinary -> LargeUtf8
5. to_base64(String) -> String -> to_base64(Binary) -> String
6. from_base64(String) -> String -> from_base64(String) -> Binary
7. to_hex(String) -> String -> to_hex(Binary) -> String
8. from_hex(String) -> String -> from_hex(String) -> Binary
9. Add utf8 check when: (a) {Binary, LargeBinary, FixLength}Array -> Column::String (2) native format read DataType::String

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 

[sundy-li]

tests failed in py-binding

[andylokandy]

https://github.com/datafuselabs/databend/blob/3709d5f1a01a7c580d732da82934b34a0bd6cd6d/tests/sqllogictests/suites/stage/formats/parquet/parquet_field_types.test#L174-L204

Are these appropriate behaviors? @b41sh @sundy-li

PS:

mysql> select * from infer_schema(location=>'@data/parquet/json_string.parquet');
+-------------+--------+----------+----------+
| column_name | type   | nullable | order_id |
+-------------+--------+----------+----------+
| a           | INT    |        1 |        0 |
| b           | BINARY |        1 |        1 |
+-------------+--------+----------+----------+
2 rows in set (0.02 sec)

mysql> select * from @data/parquet/json_string.parquet;
+------+----------------------------------------------------+
| a    | b                                                  |
+------+----------------------------------------------------+
|    1 | 7B226B31223A227631222C226B32223A227632227D         |
|    2 | 7B226B223A7B226B31223A3132332C226B32223A3435367D7D |
|    3 | 5B312C322C335D                                     |
|    4 | 5B2261222C2262225D                                 |
+------+----------------------------------------------------+

mysql> select typeof(b) from @data/parquet/json_string.parquet;
+-------------+
| typeof(b)   |
+-------------+
| BINARY NULL |
| BINARY NULL |
| BINARY NULL |
| BINARY NULL |
+-------------+

[sundy-li]

select * from @data/parquet/json_string.parquet;

Schema is binary because parquet did not have variant datatype, so we use binary to store variant.
Since we are reading from parquet, it's correct to show as binary.

But if the parquet file is inside fuse engine, it'll show as JSON since we have variant datatype.

[andylokandy]

It's ready for review