Accelerate Iceberg when reading partition columns only

[findepi]

Description

Avoid data files I/O when

• reading only partitioning columns
• doing count(*) queries on Iceberg tables (with no group by, or grouping by partitioning columns)

[findepi]

cc @osscm

[findinpath]

Please add tests when doing count(*) from the whole table.

Add a test involving a filter on a non partition column on a partitioned table.

Also important check the file system accesses for min(C)/max(C) .

Also add another test when there are delete files for the table.

[maswin]

We have a query which joins with a table only on partition columns.
Due to this part, we are not able to utilize the combined table scan.

if (wholeFileTask.deletes().isEmpty() && noDataColumnsProjected(wholeFileTask)) {
    fileTasksIterator = List.of(wholeFileTask).iterator();
}
else {
    fileTasksIterator = wholeFileTask.split(targetSplitSize).iterator();
}

The table has 1 million small files, but it is creating more than 1 million splits since it is not converting this scan into SplittableScanTask
Any reason why we are skipping combining splits when no data columns are projected?