GroupedHashAggregateStream should create smaller spill batches

[milenkovicm]

Is your feature request related to a problem or challenge?

At the moment GroupedHashAggregateStream will spill state as a single batch, which is not optimal when merging as it loads whole file in memory as a single batch.

Describe the solution you'd like

I'd like to spit spill batch into smaller chunks with default batch size same as default batch size set in configuration property.

Describe alternatives you've considered

I have considered setting batch size to a fixed size or read from configuration property, but at the moment I did not do it as it would be bigger change.

Additional context

Relates to #7858