[TopNOperator-Spilling] [Part1] Move GroupByHash management inside GroupedTopNBuilder

[shrinidhijoshi]

This PR is Part 1 of 3 part PR to implement Spilling in TopNOperator/TopNRowNumberOperator Part2, Part3)

In this PR we encapsulate the GroupByHash inside the InMemoryGroupedTopNBuilder so that it is easier to have multiple instances of GroupedTopNBuilder in the same class

Test plan - Unit tests show now failures

== NO RELEASE NOTE ==

[highker]

If you want, feel free to schedule a design review. If you are only focusing on TopNOperator, the workflow should be fairly simple. If you want to attack TopNRowNumberOperator, things will be a lot more complicated.

For TopNOperator, ideally, the process would be:

1. having a spiller in the operator
2. once the memory hits the limit, drain the GroupedTopNBuilder and spill
3. create a new GroupedTopNBuilder and repeat
4. once the operator finishes, do external merge.

The process would be fairly similar to OrderByOperator

[highker]

It's anti pattern to pass operator-related fields into a non-operator class.... We probably don't need to change anything in this class. It's up to the calling operators to spill. OrderByOperator has the closest logic.

[shrinidhijoshi]

Ah that's a good point. I borrowed this pattern from SpillableHashAggregationBuilder as I was implementing for TopNRowNumberOperator as well, which does grouped TopN and in that aspect SpillableHashAggregationBuilder is closer to this problem. I can reference OrderByOperator for code patterns.

[shrinidhijoshi]

Update from offline discussions:
Instead of moving GroupByHash into the GroupedTopNBuilder, we can create a GroupByHash Supplier/Factory and in the Operator pass it to the TopNBuilder.
This will avoid

1. leaking operator related fields (operatorContext, user/revocableMemoryContext, etc) and logic into TopNBuilder
2. Avoid functions with long list of argument fields

[shrinidhijoshi]

@highker Thanks for the review. Yes I can schedule a design review meeting. I agree that for TopNOperator its fairly simple. But I am solving for both TopNOperator and TopnRowNumber operator, as can be seen in the #18400

[shrinidhijoshi]

As a part of the design review. We also discussed that we don't need to split the PR into 3 parts. So will be closing this PR