[native_iceberg_compat] Spark SQL core-2 "filter pushdown - StringPredicate" failure

[andygrove]

Describe the bug

Test fails:

[info] - filter pushdown - StringPredicate *** FAILED *** (3 seconds, 333 milliseconds)
[info]   0 was not greater than 0 (ParquetFilterSuite.scala:242)

Debug logging shows the filter that should have been pushed down:

testStringPredicate filter=value like 'a%' shouldFilterOut=true

Comet plan shows that the filter is pushed down:

*(1) CometColumnarToRow
+- CometFilter [value#14], (isnotnull(value#14) AND StartsWith(value#14, a))
   +- CometScan parquet [value#14] Batched: true, DataFilters: [isnotnull(value#14), StartsWith(value#14, a)], Format: CometParquet, Location: InMemoryFileIndex(1 paths)[file:/Users/andy/git/apache/apache-spark/target/tmp/spark-e8f77490-5e9..., PartitionFilters: [], PushedFilters: [IsNotNull(value), StringStartsWith(value,a)], ReadSchema: struct<value:string>

However, this fails to filter out partitions during Spark execution.

Steps to reproduce

No response

Expected behavior

No response

Additional context

No response