[FEA] Support rank as window function

[sperlingxx]

Is your feature request related to a problem? Please describe.
rank is a frequently used window function within SQL context, same as: row_number or count. Spark catalyst supports rank as window function, but spark-rapids hasn't provided GPU overriding of rank.

Describe the solution you'd like
Like row_number or other window functions, GPU overriding of rank should be relied on cuDF rolling window aggregation. However, cuDF has yet supported rank as a kind of aggregation. So, we need to support rank in cuDF in primary. Referring the implementation of row_number, we only needs to implement process_rolling_window function for rank if we don't support rank with time_range_rolling_window (which could be done in future).

[sameerz]

Depends on rapidsai/cudf#8138 , rapidsai/cudf#8158

[sameerz]

Depends on rapidsai/cudf#8652