[Fix](Nereids) Group by binding should be consistent with the behavior of the old optimizer

[zhengshiJ]

Proposed changes

Issue Number: close #15328

Problem summary

The behavior of group by binding is consistent with the old optimizer, and it will be changed to the SQL standard in the future. It will be obtained from the child first, and then obtained from the output of the operator.

Checklist(Required)

1. Does it affect the original behavior:
   • Yes
   • No
   • I don't know
2. Has unit tests been added:
   • Yes
   • No
   • No Need
3. Has document been added or modified:
   • Yes
   • No
   • No Need
4. Does it need to update dependencies:
   • Yes
   • No
5. Are there any changes that cannot be rolled back:
   • Yes (If Yes, please explain WHY)
   • No

Further comments

If this is a relatively large or complex change, kick off the discussion at dev@doris.apache.org by explaining why you chose the solution you did and what alternatives you considered, etc...

[morrySnow]

need rebase to newest master to pass FE UT

[github-actions]

PR approved by at least one committer and no changes requested.

[github-actions]

PR approved by anyone and no changes requested.

[hello-stephen]

TeamCity pipeline, clickbench performance test result:
the sum of best hot time: 35.99 seconds
load time: 658 seconds
storage size: 17123121728 Bytes
https://doris-community-test-1308700295.cos.ap-hongkong.myqcloud.com/tmp/20221230103032_clickbench_pr_71605.html

[morrySnow]

according #14077, group by and grouping sets binding slot in same way. first use child's output. if failed, use aggregate's output. so, we do not need to change the bind order at all.
I close this PR, and will fix bugs in binding slot on aggregate in another PR.