Implement sc.get.summarized_expression_df

[gokceneraslan]

This pr moves the (in)famous _prepare_dataframe function from scanpy.plotting._anndata to sc.get as _indexed_expression_df (bcs why would it be in plotting anyway) and implements a simple public interface called sc.get.summarized_expresion_df which simply provides nonzero mean/var and fraction using _indexed_expression_df function.

As discussed here (https://github.com/theislab/scanpy/pull/1388#issuecomment-678739734) we can use this in rank_genes_groups_df.