[FEA] Support passing scalar args to df.apply

[randerzander]

I often want to use a "UDF" on a DataFrame where some scalar values can be supplied programmatically.

In Pandas, I can do:

import pandas as pd

pdf = pd.DataFrame({'val': [0, 1, 2]})

def func(x, A, B):
  return x['val'] + A - B

pdf.apply(func, axis=1, args=(1, 2))
0   -1
1    0
2    1

With cuDF, I can't see a way to pass scalar args to func:

import cudf

cdf = cudf.from_pandas(pdf)
cdf.apply(func, axis=1, args=(1, 2))
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/rgelhausen/conda/envs/dsql/lib/python3.8/site-packages/cudf-21.12.0a0+252.ga694162f09-py3.8-linux-x86_64.egg/cudf/core/dataframe.py", line 4237, in apply
    raise ValueError("args and kwargs are not yet supported.")
ValueError: args and kwargs are not yet supported.

As a workaround, I can do something like:

A = 1
B = 2

def func(x):
  return x['val'] + A - B

cdf.apply(func)

And if I want different values for A and B, I can update them then apply func again.