[core] System files table distributed execute remold

[leaves12138]

Purpose

Flink search files table takes too long, files collect by single executor may be out-of-memory too.
It's better to execute distributedly.

Tests

API and Format

Documentation

[JingsongLi]

Use:

SnapshotReader snapshotReader = storeTable.newSnapshotReader();
if (partitionPredicate != null) {
    List<BinaryRow> partitions = snapshotReader.withPartitionFilter(partitionPredicate).partitions();
}

[JingsongLi]

+1