plan: optimize the performance of decodePlan for big union query (#18891)

[ti-srebot]

cherry-pick #18891 to release-4.0

---

What problem does this PR solve?

For big query like:

select count(1) as num,a from t1 where a=1 group by a 
union 
select count(1) as num,a from t1 where a=3 group by a;
union 
select count(1) as num,a from t1 where a=5 group by a;
union
... -- many union statement

The performance of DecodePlan is bad, this PR try to optimize the performance of it:

For union with 50000 select statements:

--before this PR;

DecodePlan spend 1.35 S

-- this PR:

DecodePlan spend 84.731694ms

Related changes

• Need to cherry-pick to the release branch

Check List

Tests

• Unit test

Side effects

• No

Release note

• optimize the performance of decodePlan for big union query

[ti-srebot]

/run-all-tests

[crazycs520]

/rebuild

[crazycs520]

/run-all-tests

[wjhuang2016]

LGTM

[tangenta]

LGTM

[zz-jason]

/merge

[ti-srebot]

/run-all-tests