Add compute_global_memory_aggs used by shared memory groupby

[PointKernel]

Description

This work is part of splitting the original bulk shared memory groupby PR #16619.

When the aggregation request requires more shared memory than available and involves dictionary aggregation, we need to perform the aggregation directly in global memory using compute_global_memory_aggs.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[PointKernel]

Just realized this cannot be a standalone PR since it requires the new create_sparse_results_table which will breaks the current code.