Is your feature request related to a problem? Please describe.
In https://github.com/NVIDIA/spark-rapids/pull/12354/files#r2015510439 @abellina required to solve the can-not-spill issue for non-kudo input case for AsymmetricJoinSizer, once #12354 is checked in. To address this comment, we need to make non-kudo input for AsymmetricJoinSizer also spillable, so that merging #12354 (even though itself is fixing a bug) will increase the risk of CPU OOM