[core] Add tests and implement withFilter for StatisticTable

[heye1005]

Purpose

Remove the two // TODO markers on StatisticTable

Tests

Adds StatisticTableTest

[heye1005]

CI Failures

These appear to be unrelated flaky tests, not caused by this PR's changes:

• MySqlSyncTableActionITCase.testAllTypes (timeout)
• KafkaCanalSyncDatabaseActionITCase.testExpressionNow (timeout)
• Maven Central network timeout
  Could you help re-run the failed jobs?

I also want to separate new pr to address the CDC ITCase flakiness.

[JingsongLi]

Review comment test

[JingsongLi]

Review: StatisticTable withFilter implementation

Overall this is a clean, well-structured change that resolves the TODO markers and adds solid test coverage. A few observations:

Correctness

Filter-before-projection ordering is correct. The postFilter is applied to the full-schema row produced by toRow(), and projection via ProjectedRow happens afterward. This ensures predicate column indices align with TABLE_TYPE rather than the projected readType.

Scan-level withFilter is intentionally a no-op -- since StatisticTable always produces at most one split, there is nothing to prune at the scan layer. The actual row-level filtering is delegated to StatisticRead.withFilter, which is the right approach for this table.

Minor suggestion

The Iterators.filter(rows, postFilter::test) call: since Statistics yields at most one row, the practical effect of the filter is a simple keep-or-discard decision. The current implementation is fine and consistent with how other system tables handle filtering.

Tests

Good coverage: empty statistics, basic read, predicate hit/miss on snapshot_id, and range predicate on merged_record_count. One optional enhancement for the future: a test combining withFilter and withProjection together to validate the filter-then-project ordering invariant.

Summary

The change is correct and well-tested. LGTM.

[JingsongLi]

+1