[fix](agg) Aggregating string types with null values may result in incorrect result (#42067)

Aggregating string types with null values may result in incorrect result
because using the replace_column_data function can cause incorrect
offsets in the column.

A reproducible case：
```
CREATE TABLE `test_scan_keys_with_bool_type` (
    `col1` tinyint NOT NULL,
    `col2` int NOT NULL,
    `col3` tinyint NOT NULL,
    `col5` boolean REPLACE NOT NULL,
    `col4` datetime(2) REPLACE NOT NULL,
    `col6` double REPLACE_IF_NOT_NULL NULL,
    `col7` varchar(100) REPLACE_IF_NOT_NULL NULL
) ENGINE=OLAP
AGGREGATE KEY(`col1`, `col2`, `col3`)
DISTRIBUTED BY HASH(`col1`, `col2`, `col3`) BUCKETS 1
PROPERTIES (
    "replication_allocation" = "tag.location.default: 1",
    "disable_auto_compaction" = "true"
);

insert into test_scan_keys_with_bool_type values
( -100 ,    1 ,  -82 ,    1 , '2024-02-16 04:37:37.00' , -1299962421.904282 , NULL ),
( -100 ,    0 ,  -82 ,    1 , '2024-02-16 04:37:37.00' , -1299962421.904282 , "hi" ),
( -100 ,    1 ,   92 ,    1 , '2024-02-16 04:37:37.00' ,   23423423.0324234 , NULL );

 insert into test_scan_keys_with_bool_type values
( -100 ,    1 ,  1 ,    1 , '2024-02-16 04:37:37.00' , -1299962421.904282 , "doris" );

MySQL [test]> select * from test_scan_keys_with_bool_type;
+------+------+------+------+------------------------+---------------------+-------+
| col1 | col2 | col3 | col5 | col4                   | col6                | col7  |
+------+------+------+------+------------------------+---------------------+-------+
| -100 |    0 |  -82 |    1 | 2024-02-16 04:37:37.00 | -1299962421.9042821 | hi    |
| -100 |    1 |  -82 |    1 | 2024-02-16 04:37:37.00 | -1299962421.9042821 | NULL  |
| -100 |    1 |    1 |    1 | 2024-02-16 04:37:37.00 | -1299962421.9042821 | hidor |
| -100 |    1 |   92 |    1 | 2024-02-16 04:37:37.00 |    23423423.0324234 | NULL  |
+------+------+------+------+------------------------+---------------------+-------+
4 rows in set (0.057 sec)

MySQL [test]> set skip_storage_engine_merge = true; select * from test_scan_keys_with_bool_type;
+------+------+------+------+------------------------+---------------------+-------+
| col1 | col2 | col3 | col5 | col4                   | col6                | col7  |
+------+------+------+------+------------------------+---------------------+-------+
| -100 |    0 |  -82 |    1 | 2024-02-16 04:37:37.00 | -1299962421.9042821 | hi    |
| -100 |    1 |  -82 |    1 | 2024-02-16 04:37:37.00 | -1299962421.9042821 | NULL  |
| -100 |    1 |   92 |    1 | 2024-02-16 04:37:37.00 |    23423423.0324234 | NULL  |
| -100 |    1 |    1 |    1 | 2024-02-16 04:37:37.00 | -1299962421.9042821 | doris |
+------+------+------+------+------------------------+---------------------+-------+
4 rows in set (0.023 sec)
```
#33493 By supporting variant type
aggregation, this issue has been resolved.So versions after 2.1 do not
have this issue.

Author: liaoxin01

be/src/vec/olap/block_reader.cpp

@@ -470,10 +470,10 @@ size_t BlockReader::_copy_agg_data() {
         auto& dst_column = _stored_data_columns[idx];
         if (_stored_has_variable_length_tag[idx]) {
             //variable length type should replace ordered
+            dst_column->clear();
             for (size_t i = 0; i < copy_size; i++) {
                 auto& ref = _stored_row_ref[i];
-                dst_column->replace_column_data(*ref.block->get_by_position(idx).column,
-                                                ref.row_pos, i);
+                dst_column->insert_from(*ref.block->get_by_position(idx).column, ref.row_pos);
             }
         } else {
             for (auto& it : _temp_ref_map) {

regression-test/data/data_model_p0/aggregate/test_aggregate_table.out

@@ -46,3 +46,9 @@ datetimev2_value_min_2	datetime(6)	Yes	false	\N	MIN
 datetimev2_value_replace_2	datetime(6)	Yes	false	\N	REPLACE
 datetimev2_value_replace_if_not_null_2	datetime(6)	Yes	false	\N	REPLACE_IF_NOT_NULL
 
+-- !string_agg_table_with_null --
+-100	0	-82	true	2024-02-16T04:37:37	-1.299962421904282E9	hi
+-100	1	-82	true	2024-02-16T04:37:37	-1.299962421904282E9	\N
+-100	1	1	true	2024-02-16T04:37:37	1.399962421904282E9	doris
+-100	1	92	true	2024-02-16T04:37:37	2.34234230324234E7	NULL
+

regression-test/suites/data_model_p0/aggregate/test_aggregate_table.groovy

@@ -99,4 +99,36 @@ suite("test_aggregate_table") {
     qt_desc_date_table """desc date_agg"""
     sql """DROP TABLE date_agg"""
 
+    sql """DROP TABLE IF EXISTS test_string_agg_with_null"""
+    sql """ 
+        CREATE TABLE `test_string_agg_with_null` (
+            `col1` tinyint NOT NULL,
+            `col2` int NOT NULL,
+            `col3` tinyint NOT NULL,
+            `col5` boolean REPLACE NOT NULL,
+            `col4` datetime(2) REPLACE NOT NULL,
+            `col6` double REPLACE_IF_NOT_NULL NULL,
+            `col7` varchar(100) REPLACE_IF_NOT_NULL NULL
+        ) ENGINE=OLAP
+        AGGREGATE KEY(`col1`, `col2`, `col3`)
+        DISTRIBUTED BY HASH(`col1`, `col2`, `col3`) BUCKETS 1
+        PROPERTIES (
+            "replication_allocation" = "tag.location.default: 1",
+            "disable_auto_compaction" = "true"
+        ); 
+    """
+
+    sql """ insert into test_string_agg_with_null values
+        ( -100 ,    1 ,  -82 ,    1 , '2024-02-16 04:37:37.00' , -1299962421.904282 , NULL ),
+        ( -100 ,    0 ,  -82 ,    1 , '2024-02-16 04:37:37.00' , -1299962421.904282 , "hi" ),
+        ( -100 ,    1 ,   92 ,    1 , '2024-02-16 04:37:37.00' ,   23423423.0324234 , "NULL" );
+    """
+
+    sql """ insert into test_string_agg_with_null values
+        ( -100 ,    1 ,  1 ,    1 , '2024-02-16 04:37:37.00' , 1399962421.904282 , "doris" );
+        """
+    
+    qt_string_agg_table_with_null """ select * from test_string_agg_with_null """
+    sql """DROP TABLE test_string_agg_with_null"""
+
 }