[FEA] DataFrame and Index nunique

[beckernick]

In Pandas, DataFrame.nunique returns Series in which each row index corresponds to a column name in the original dataframe and the value corresponds to the number of distinct elements in that column. The Index methods behave like the Series method.

We already have the machinery to run this with {DataFrame, SingleColumnFrame}._reduce, but it would be nice to expose this as a top level API on DataFrame (and Index). The Series and Groupby APIs are already implemented.

import pandas as pd
​
df = pd.DataFrame({
    "key": [0, 1, 1, 0, 0, 1],
    "val": [1, 8, 3, 9, -3, 8],
})
print(df, "\n")
print(df.nunique())
   key  val
0    0    1
1    1    8
2    1    3
3    0    9
4    0   -3
5    1    8 

key    2
val    5
dtype: int64

import cudf
​
df = cudf.DataFrame({
    "key": [0, 1, 1, 0, 0, 1],
    "val": [1, 8, 3, 9, -3, 8],
})
print(df, "\n")
print(df.index._reduce("distinct_count"))
print(df.key._reduce("distinct_count"))
   key  val
0    0    1
1    1    8
2    1    3
3    0    9
4    0   -3
5    1    8 

6
2