Physical planning: cast low-medium cardinality columns to dictionary arrays before aggregation

[Rich-T-kid]

Is your feature request related to a problem or challenge?

GroupValuesDictionary can provide up to a 2-5x speedup over GroupValuesRows for aggregations, but it is only used when the input data is already a dictionary-encoded column.

Describe the solution you'd like

A physical optimizer rule could inspect column statistics at plan time and cast low-medium cardinality Utf8 columns to Dictionary(Int32, Utf8) before aggregation, allowing GroupValuesDictionary to be selected automatically without any changes from the user.
Parquet files keep metadata about cardinality, this may be a good place to start

Describe alternatives you've considered

No response

Additional context

#21765

[Rich-T-kid]

this won't be a priority until #21878 is complete

[Rich-T-kid]

This may still have use for single string columns -> dictionary column for aggregations