[VL]When reading a decimal type with large precision from an Iceberg-Parquet table, the data is mismatch.

### Backend

VL (Velox)

### Bug description

When I read an Iceberg table with a column of type decimal(28,8) and the value is 0, the result returned by Gluten is 184467438628.42748416.

### Gluten version

_No response_

### Spark version

Spark-3.3.x

### Spark configurations

_No response_

### System information

_No response_

### Relevant logs

```bash

```

Start by reproducing the decimal(28,8) read from an Iceberg-Parquet table using Spark 3.3.x through the VL backend, with a value of 0. Compare the returned value with the expected zero and use the relevant logs and environment details to isolate the decimal conversion path; done means the read returns the correct value.

Written by the indexing model from the issue text.