[Improvement] Better support for AQE sortMergeJoin -> broadcastHashJoin

### Code of Conduct

- [X] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct)

### Search before asking

- [X] I have searched in the [issues](https://github.com/apache/incubator-uniffle/issues?q=is%3Aissue) and found no similar issues.

### What would you like to be improved?

In #630 , we disable the `localShuffleRead` for AQE in spark. For a pity, sometimes the partial partition is so huge, but the broadcastHashJoin won't optimize skew join. This optimization will be solved by spark community ASSP.

After that, I will propose a new patch for uniffle to better support Spark older version.

### How should we improve?

_No response_

### Are you willing to submit PR?

- [X] Yes I am willing to submit a PR!

Start by reading issue #630 and reviewing the Spark AQE behavior described here, especially the transition from sort-merge join to broadcast-hash join. The work is done when Uniffle has the proposed improved support for older Spark versions without disabling the needed optimization.

Written by the indexing model from the issue text.