[CH] Wired type result of decimal plus

### Backend

CH (ClickHouse)

### Bug description

`select typeof((p*(1.0-p)/n+z*z/(4.0*n*n))) from (select cast('0E-14' as decimal(38,14)) as p, cast('1.96' as decimal(3,2)) as z, cast(37 as long) as n, id from range(1));`
got:
```
+----------------------------------------------------+
| typeof((((p * (1.0 - p)) / n) + ((z * z) / ((4.0 * n) * n)))) |
+----------------------------------------------------+
| decimal(38,6) |
```

`select typeof( (z*z/(4.0*n*n))) from (select cast('0E-14' as decimal(38,14)) as p, cast('1.96' as decimal(3,2)) as z, cast(37 as long) as n, id from range(1));`
got:
```
+--------------------------------------+
| typeof(((z * z) / ((4.0 * n) * n))) |
+--------------------------------------+
| decimal(38,34) |
```

`select typeof( (p*(1.0-p)/n)) from (select cast('0E-14' as decimal(38,14)) as p, cast('1.96' as decimal(3,2)) as z, cast(37 as long) as n, id from range(1));`
got:
```
+--------------------------------+
| typeof(((p * (1.0 - p)) / n)) |
+--------------------------------+
| decimal(38,6) |
```

### Spark version

Spark-3.3.x

### Spark configurations

_No response_

### System information

_No response_

### Relevant logs

```bash

```

Start by running the supplied Spark 3.3.x query against the ClickHouse backend and compare the reported decimal types with Spark's expected behavior. Trace the decimal type inference used for arithmetic expressions; done means the expression reports the intended precision and scale and the regression is covered by an appropriate test.

Written by the indexing model from the issue text.