[FEA] Implement merge for dataframes with decimal columns

[ChrisJar]

Is your feature request related to a problem? Please describe.
I would like to be able to merge dataframes that contain columns with type decimal

Describe the solution you'd like
I would like to merge DataFrames with decimal columns the way I can currently merge DataFrames with float columns. For example:

df1 = cudf.DataFrame({'id': [0, 1, 1], 'val': [1.00, 1.01, 1.02]})
df2 = cudf.DataFrame({'id': [0, 1, 1], 'val': [8.28, 9.32, 4.94]})
df1.merge(df2, left_on=['id'], right_on=['id'], how='inner')

returns

	id	val_x	val_y
0	0	1.00	9.32
1	1	1.01	8.28
2	1	1.02	8.28
3	1	1.01	4.94
4	1	1.02	4.94

However, after casting the float columns to decimal columns

df1['val'] = cudf.Series([decimal.Decimal(x) for x in [1.00, 1.01, 1.02]], dtype=cudf.Decimal64Dtype(7,3))
df2['val'] = cudf.Series([decimal.Decimal(x) for x in [8.28, 9.32, 4.94]], dtype=cudf.Decimal64Dtype(7,3))
df1.merge(df2, left_on=['id'], right_on=['id'], how='inner')

It returns

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-19-6a6f615495ed> in <module>
----> 1 df1.merge(df2, left_on=['id'], right_on=['id'], how='inner')

/home/u00u7rh1e72hXfsipJ357/miniconda3/envs/rapids-gpu-bdb/lib/python3.7/contextlib.py in inner(*args, **kwds)
     72         def inner(*args, **kwds):
     73             with self._recreate_cm():
---> 74                 return func(*args, **kwds)
     75         return inner
     76 

/home/u00u7rh1e72hXfsipJ357/miniconda3/envs/rapids-gpu-bdb/lib/python3.7/site-packages/cudf/core/dataframe.py in merge(self, right, on, left_on, right_on, left_index, right_index, how, sort, lsuffix, rsuffix, method, indicator, suffixes)
   4239             method=method,
   4240             indicator=indicator,
-> 4241             suffixes=suffixes,
   4242         )
   4243         return gdf_result

/home/u00u7rh1e72hXfsipJ357/miniconda3/envs/rapids-gpu-bdb/lib/python3.7/site-packages/cudf/core/frame.py in _merge(self, right, on, left_on, right_on, left_index, right_index, how, sort, lsuffix, rsuffix, method, indicator, suffixes)
   3434             suffixes,
   3435         )
-> 3436         to_return = mergeop.perform_merge()
   3437 
   3438         # If sort=True, Pandas would sort on the key columns in the

/home/u00u7rh1e72hXfsipJ357/miniconda3/envs/rapids-gpu-bdb/lib/python3.7/site-packages/cudf/core/join/join.py in perform_merge(self)
    119         )
    120         result = self.out_class._from_table(libcudf_result)
--> 121         result = self.typecast_libcudf_to_output(result, output_dtypes)
    122         if isinstance(result, cudf.Index):
    123             return result

/home/u00u7rh1e72hXfsipJ357/miniconda3/envs/rapids-gpu-bdb/lib/python3.7/site-packages/cudf/core/join/join.py in typecast_libcudf_to_output(self, output, output_dtypes)
    412             if data_dtype:
    413                 output._data[data_col_lbl] = self._build_output_col(
--> 414                     data_col, data_dtype
    415                 )
    416         return output

/home/u00u7rh1e72hXfsipJ357/miniconda3/envs/rapids-gpu-bdb/lib/python3.7/site-packages/cudf/core/join/join.py in _build_output_col(self, col, dtype)
    427             )
    428         else:
--> 429             outcol = col.astype(dtype)
    430         return outcol

/home/u00u7rh1e72hXfsipJ357/miniconda3/envs/rapids-gpu-bdb/lib/python3.7/site-packages/cudf/core/column/column.py in astype(self, dtype, **kwargs)
   1027                 )
   1028             return self
-> 1029         elif np.issubdtype(dtype, np.datetime64):
   1030             return self.as_datetime_column(dtype, **kwargs)
   1031         elif np.issubdtype(dtype, np.timedelta64):

/home/u00u7rh1e72hXfsipJ357/miniconda3/envs/rapids-gpu-bdb/lib/python3.7/site-packages/numpy/core/numerictypes.py in issubdtype(arg1, arg2)
    386     """
    387     if not issubclass_(arg1, generic):
--> 388         arg1 = dtype(arg1).type
    389     if not issubclass_(arg2, generic):
    390         arg2 = dtype(arg2).type

TypeError: Cannot interpret 'Decimal64Dtype(precision=7, scale=3)' as a data type

[harrism]

This already works in libcudf, right @codereport ?

[codereport]

This already works in libcudf, right @codereport ?

Yes, libcudf supports this.