MAINT: Adapt numerical promotion to NumPy 2 and Pandas 2.2

Pandas keeps using weak promotion even for strongly typed
"scalars" (i.e. 0-d objects).
This tries to (mostly) match that, but there may be better ways to
do it.  I am having difficulty to think of the best way though.

python/cudf/cudf/core/column/numerical.py

@@ -291,15 +291,28 @@ def normalize_binop_value(
         if isinstance(other, cudf.Scalar):
             if self.dtype == other.dtype:
                 return other
+
             # expensive device-host transfer just to
             # adjust the dtype
             other = other.value
+
+            # NumPy 2 needs a Python scalar to do weak promotion, but
+            # pandas forces weak promotion always
+            # TODO: We could use 0, 0.0, and 0j for promotion to avoid copies.
+            if other.dtype.kind in "ifc":
+                other = other.item()
+        elif not isinstance(other, (int, float, complex)):
+            # Go via NumPy to get the value
+            other = np.array(other)
+            if other.dtype.kind in "ifc":
+                other = other.item()
+
         # Try and match pandas and hence numpy. Deduce the common
-        # dtype via the _value_ of other, and the dtype of self. TODO:
-        # When NEP50 is accepted, this might want changed or
-        # simplified.
-        # This is not at all simple:
-        # np.result_type(np.int64(0), np.uint8)
+        # dtype via the _value_ of other, and the dtype of self on NumPy 1.x
+        # with NumPy 2, we force weak promotion even for our/NumPy scalars
+        # to match pandas 2.2.
+        # Weak promotion is not at all simple:
+        # np.result_type(0, np.uint8)
         #   => np.uint8
         # np.result_type(np.asarray([0], dtype=np.int64), np.uint8)
         #   => np.int64