[FEA] Add Python bindings for drop_list_duplicates

[argenisleon]

Is your feature request related to a problem? Please describe.
I wish I could deduplicate the elements in a column list. For example, applying the function to this dataframe:
0 [optimus, a, optimus]
1 [optimus, b]
2 [optimus, c]
Name: A, dtype: list

should output
0 [optimus, a]
1 [optimus, b]
2 [optimus, c]
Name: A, dtype: list

Describe the solution you'd like
In pandas you could use someting like df.apply(lambda x: set(x['column_list']), axis=1). No sure which could be the best approach in cudf

Describe alternatives you've considered

text = ["Optimus a", 
        "Optimus b Optimus",
       "Optims c"]
df = cudf.DataFrame({'text':text,'doc_id':[0,1,2]})
tokens = df.text.str.tokenize(' ')
tk_cnts = df.text.str.token_count(' ')
global_pos = df.doc_id.repeat(tk_cnts).reset_index(drop=True)

tokenized_df = cudf.DataFrame({'token':tokens,'doc_id':global_pos})
tokenized_df['token_order']=1
tokenized_df['token_order'] = tokenized_df['token_order'].cumsum()

tokenized_df.columns = ['token_ser', 'output_column_id_ser', 'ouput_token_order_ser']
tokenized_df= tokenized_df.drop_duplicates(["token_ser","output_column_id_ser"])
tokenized_df["token_ser"] = tokenized_df["token_ser"].str.detokenize('output_column_id_ser')

[kkraus14]

I think this would make sense to expose as something like cudf.Series.list.unique() in cuDF Python

[ttnghia]

We have a new PR for drop_list_duplicates in cudf C++: #7528
Please work on the python side if necessary. Thanks.