`native_datafusion` reports twice more files and bytes scanned

[comphead]

Describe the bug

spark.range(10000000).repartition(20).write.parquet(location) 
spark.read.parquet("location").show(false)

which are twice more than original. The issue seems to be for metric only, the actual file read only once

Steps to reproduce

No response

Expected behavior

No response

Additional context

No response

[0lai0]

take