[fix](nereids) fix union all instance number

[xzj7019]

In following plan pattern, fragment 11 has 3 child fragments aligning with 3 exchange nodes and the union plan node only have 2 plan children, i.e, the aggr node and the exchange node(4). In current union node's parallelism computing, it used the plan children's size to do iteration to find the max parallel but ignore the last fragment's. Actually, it should use fragment's children(exclude unpartitioned ones) instead of plan node's children to do the iteration.

11:VUNION(703)
|
|----10:VAGGREGATE (update finalize)(697)
| | group by: brand_name[#23], skc[#24]
| | sortByGroupKey:false
| | cardinality=1
| | final projections: '1', brand_name[#25], skc[#26]
| | final project output tuple id: 11
| | distribute expr lists: skc[#24]
| |
| 9:VHASH JOIN(691)
| | join op: INNER JOIN(PARTITIONED)[]
| | equal join conjunct: (skc[#19] = skc_code[#12])
| | cardinality=1
| | vec output tuple id: 9
| | output tuple id: 9
| | vIntermediate tuple ids: 8
| | hash output slot ids: 18 19
| | final projections: brand_name[#20], skc[#21]
| | final project output tuple id: 9
| | distribute expr lists: skc[#19]
| | distribute expr lists: skc_code[#12]
| |
| |----6:VEXCHANGE
| | offset: 0
| | distribute expr lists: skc_code[#12]
| |
| 8:VEXCHANGE
| offset: 0
| distribute expr lists:
|
4:VEXCHANGE
offset: 0
distribute expr lists:

[xzj7019]

run buildall

[xzj7019]

run buildall

[github-actions]

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

[github-actions]

PR approved by anyone and no changes requested.