[FEA] Refactor aggregation APIs

[ttnghia]

Currently, aggregation APIs (groupby, reductions, rolling, etc.) are scattered around in multiple files and there are inconsistencies between the directory structures in cpp/include/, cpp/src/, cpp/tests/, and cpp/benchmarks/. For example:

cpp/include/:

• include/cudf/aggregation.hpp
• include/cudf/groupby.hpp
• include/cudf/rolling.hpp
• ....

cpp/src/:

• src/aggregation/.
• src/groupby/.
• src/rolling/.
• ....

cpp/tests/:

• tests/aggregation/.
• tests/groupby/.
• tests/grouped_rolling/.
• tests/lead_lag/.
• ....

I suggest to group all aggregation APIs together, so those APIs will be in include/aggreation/*.*, src/aggregation/*.*, tests/aggregation/*.* etc.

[harrism]

I think this should be covered as part of #7106 and #6895

[jrhemstad]

I don't think it makes sense to group everything that touches aggregation under a aggregation/ directory. It's too broad a category to be useful. Grouping groupby,reduction,scan,rolling,etc. into the same directory is more harmful than helpful.

[ttnghia]

Yes, it would be complex to combine everything to one. So, how about make them sub-categories? For example, we can have

aggregation/grouby/
aggregation/reduction/
aggregation/rolling/

Grouping the related APIs together make things easier to access and refer to.

[jrhemstad]

I still think that hurts more than it helps. The fact that a very diverse set of algorithms share a common aggregation input parameter is an implementation detail. Adding another layer of indirection to group things by an implementation detail doesn't seem worth it.

[harrism]

Good point. Moreover, not all groupby operations are aggregations. Same goes for rolling. Suggest we close this as wontfix.