perf: merge identical agg calls in a streaming query

[stdrc]

This can increase performance in two cases:

1. select sum(a), avg(a), ... (perf: nexmark q17 #7351 (comment))
2. select count(*), ..., at the same time we internally insert a count(*) to the beginning, then we'll have plan like this:

   └─StreamAppendOnlyHashAgg { group_key: [...], aggs: [count, count, ...] }
   

Seems these two cases need different optimization logic, so track it here:

• opt(agg): reuse existing agg calls while building LogicalAgg #8196
• opt(agg): reuse existing count(*) while generating stream plan #8197