[FEA] Register Dataframe Accessors

[ericmjl]

Is your feature request related to a problem? Please describe.
I wish cuDF could allow for custom dataframe accessors. That would allow us to port over pyjanitor code to work with cuDFs as well.

What are dataframe accessors?

They are available here: https://pandas.pydata.org/docs/development/extending.html

The gist is that custom dataframe accessors allows us to do things like:

def my_custom_func(df, arg1, arg2, kwarg1, kwarg2):
    return modified_df

# Now, execute
df = df.clean_names().my_custom_func(arg1, arg2, **kwargs)

How might this help?

pyjanitor (for which I am one of the maintainers) provides a series of super convenient data cleaning functions that operate on pandas dataframes. Some of the functions make use of pandas' native API, and can be considered higher-order functions that make routine common pandas snippets. I would like to see this available for cuDF as well; doing so would help me complete the jump to using cuDF!