feat: add size-aware join selection for pipelined jobs

ballista/scheduler/src/state/aqe/execution_plan/dynamic_join.rs

@@ -825,7 +825,7 @@ const DEFAULT_BINARY_WIDTH_BYTES: usize = 100;
 ///
 /// Returns `None` if the estimate overflows, so the caller declines to broadcast
 /// rather than wrapping around to a small number.
-fn estimate_output_byte_size(
+pub(crate) fn estimate_output_byte_size(
     schema: &Schema,
     num_rows: usize,
     column_statistics: &[ColumnStatistics],

ballista/scheduler/src/state/mod.rs

@@ -50,6 +50,9 @@ pub mod execution_graph_dot;
 pub mod execution_stage;
 /// Executor registration and management.
 pub mod executor_manager;
+// Nothing plans pipelined jobs yet, so only the tests use this module.
+#[allow(dead_code)]
+pub(crate) mod pipelined_join_selection;
 /// Session state management.
 pub mod session_manager;
 /// Per-task plan rewriter (restrict scan/shuffle-reader to task's slice).