[BUG] max(BIGINT) Returns Wrong Value with High Cardinality Input and Velox Engine

### Backend

VL (Velox)

### Bug description

When running a GROUP BY query with max(ts) on a BIGINT column, Gluten 1.3 with the Velox backend returns an incorrect result for some inputs.
example：
```
select
uid,
max(msg_ts) as last_msg_ts
from
test_parquet_max
where uid = '710726744'
group by uid
```
excepted result：
uid: 710726744 last_msg_ts: 1749185469637
Actual result
uid: 710726744 last_msg_ts: 140485873811016

veloxplan：
```
I20250730 10:10:50.677613 28032 VeloxRuntime.cc:145] ############### Velox plan for task [Stage: 0 TID: 5] ###############
-- Project[3][expressions: (n3_2:INTEGER, hash_with_seed(42,"n1_5")), (n3_3:BIGINT, "n1_5"), (n3_4:BIGINT, "n2_1")] -> n3_2:INTEGER, n3_3:BIGINT, n3_4:BIGINT
-- Aggregation[2][PARTIAL [n1_5] n2_1 := max_partial("n1_4")] -> n1_5:BIGINT, n2_1:BIGINT
-- Project[1][expressions: (n1_4:BIGINT, "n0_0"), (n1_5:BIGINT, "n0_1")] -> n1_4:BIGINT, n1_5:BIGINT
-- TableScan[0][table: hive_table, remaining filter: (and(isnotnull("uid"),equalto("uid",710726744)))] -> n0_0:BIGINT, n0_1:BIGINT, n0_2:VARCHAR, n0_3:VARCHAR

I20250730 10:10:54.165143 28032 VeloxRuntime.cc:145] ############### Velox plan for task [Stage: 2 TID: 20] ###############
-- Aggregation[1][SINGLE [n0_0] n1_1 := max_merge_extract("n0_1")] -> n0_0:BIGINT, n1_1:BIGINT
-- ValueStream[0][] -> n0_0:BIGINT, n0_1:BIGINT
```

### Gluten version

Gluten-1.3

### Spark version

Spark-3.5.x

### Spark configurations

_No response_

### System information

_No response_

### Relevant logs

```bash

```

Start by reproducing the GROUP BY query with the Velox backend and inspect the Velox plan and logs emitted by VeloxRuntime.cc. Compare the partial max and max_merge_extract stages against the expected BIGINT result. Done means the query returns the correct maximum value for the reported high-cardinality input.

Written by the indexing model from the issue text.