Spark SQL test failures in native_datafusion scan

### Describe the bug

There are more than 100 failures in `core2` suite that seem to have the same root cause:

```
- Spark vectorized reader - without partition data column - select only top-level fields *** FAILED *** (317 milliseconds)
[info]   0 did not equal 1 Found 0 file sources in dataframe, but expected WrappedArray(struct<address:string>) (SchemaPruningSuite.scala:874)
```
