Support binops between float scalar to decimal column (#20199)

To remove a workaround in PDSH Q1 for cudf.pandas

```bash
~/cudf$ python $PDSH_FILE $queries --scale=0.1 --iterations 1 --path "$dataset_path"
Empty DataFrame
Columns: []
Index: []
Query 0 - Iteration 0 finished in 0.0007s
  l_returnflag l_linestatus     sum_qty  sum_base_price  sum_disc_price    sum_charge  count_order
0            A            F  3774200.00   5320753880.69   -5.054096e+09 -5.256751e+09       147790
1            N            F    95257.00    133737795.84   -1.271324e+08 -1.322863e+08         3765
2            N            O  7459297.00  10512270008.90   -9.986238e+09 -1.038558e+10       292000
3            R            F  3785523.00   5337950526.47   -5.071819e+09 -5.274406e+09       148301
Query 1 - Iteration 0 finished in 0.2097s
Iteration Summary
=======================================
query: 0
path: /home/nfs/mroeschke/cudf/sf0.1
scale_factor: 0.1
executor: in-memory
iterations: 1
---------------------------------------
min time : 0.0007
max time : 0.0007
mean time: 0.0007
=======================================
query: 1
path: /home/nfs/mroeschke/cudf/sf0.1
scale_factor: 0.1
executor: in-memory
iterations: 1
---------------------------------------
min time : 0.2097
max time : 0.2097
mean time: 0.2097
=======================================
Total mean time across all queries: 0.2103 seconds
```

Authors:
  - Matthew Roeschke (https://github.com/mroeschke)

Approvers:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

URL: #20199

Author: mroeschke

python/cudf/cudf/core/column/decimal.py

@@ -17,7 +17,7 @@
 from cudf.api.types import is_scalar
 from cudf.core._internals import binaryop
 from cudf.core.buffer import acquire_spill_lock
-from cudf.core.column.column import ColumnBase
+from cudf.core.column.column import ColumnBase, as_column
 from cudf.core.column.numerical_base import NumericalBaseColumn
 from cudf.core.dtypes import (
     Decimal32Dtype,
@@ -217,7 +217,9 @@ def _binaryop(self, other: ColumnBinaryOperand, op: str):
         if isinstance(other, ColumnBase):
             if not isinstance(other, NumericalBaseColumn):
                 return NotImplemented
-            elif other.dtype.kind in {"f", "b"}:
+            elif other.dtype.kind == "f":
+                return self.astype(other.dtype)._binaryop(other, op)
+            elif other.dtype.kind == "b":
                 raise TypeError(
                     "Decimal columns only support binary operations with "
                     "integer numerical columns."
@@ -234,6 +236,8 @@ def _binaryop(self, other: ColumnBinaryOperand, op: str):
             other_cudf_dtype = other.dtype
         elif isinstance(other, (int, Decimal)):
             other_cudf_dtype = self.dtype._from_decimal(Decimal(other))
+        elif isinstance(other, float):
+            return self._binaryop(as_column(other, length=len(self)), op)
         elif is_na_like(other):
             other = pa.scalar(None, type=cudf_dtype_to_pa_type(self.dtype))
             other_cudf_dtype = self.dtype

python/cudf/cudf/pandas/_benchmarks/pdsh.py

@@ -43,17 +43,6 @@ def q1(run_config: RunConfig) -> pd.DataFrame:
         line_item_ds = get_data(
             run_config.dataset_path, "lineitem", run_config.suffix
         )
-        # TODO: Remove float64 casting once cuDF supports subtraction between float and decimal
-        line_item_ds["l_quantity"] = line_item_ds["l_quantity"].astype(
-            "float64"
-        )
-        line_item_ds["l_extendedprice"] = line_item_ds[
-            "l_extendedprice"
-        ].astype("float64")
-        line_item_ds["l_discount"] = line_item_ds["l_discount"].astype(
-            "float64"
-        )
-        line_item_ds["l_tax"] = line_item_ds["l_tax"].astype("float64")
 
         var1 = date(1998, 9, 2)
 

python/cudf/cudf/tests/series/test_binops.py

@@ -3098,3 +3098,12 @@ def test_singleton_array(binary_op, xp):
     expect = binary_op(lhs.to_pandas(), rhs_host)
     got = binary_op(lhs, rhs_device)
     assert_eq(expect, got)
+
+
+def test_binops_float_scalar_decimal():
+    result = 1.0 - cudf.Series(
+        [decimal.Decimal("1"), decimal.Decimal("-2.5"), None],
+        dtype=cudf.Decimal32Dtype(3, 2),
+    )
+    expected = cudf.Series([0.0, -3.5, None], dtype="float64")
+    assert_eq(result, expected)