[CORE] Incorrect wholestage id in WholeStageTransformerExec

### Backend

VL (Velox)

### Bug description

In vanilla spark, the stage id of the `WholeStageCodegenExec` always starts with 1 in each SQL query

However in Gluten, the stage id is generated from a global counter, which accumulates across different queries. This number can grow very large in a long running spark session.

### Gluten version

_No response_

### Spark version

None

### Spark configurations

_No response_

### System information

_No response_

### Relevant logs

```bash

```

Start by tracing WholeStageTransformerExec and comparing its stage-id handling with WholeStageCodegenExec, the two execution classes named in the report. Reproduce the behavior across multiple SQL queries and identify how the counter is scoped; done means stage IDs follow the per-query behavior described and regression coverage verifies it.

Written by the indexing model from the issue text.