Fix overflow case and clean up some logic (#18734)

The calculation of the size in bytes for the CUDA Array Interface for pylibcudf Column objects produced from a column_view and an arbitrary owner previously ran the risk of overflow because the arithmetic was performed on int32 types but that is actually the maximum size in number of elements, not bytes. Since the CAI is a Python object, we can do the arithmetic with pure Python (infinite precision) integers to avoid this problem. In the process of fixing this bug, this PR also does some minor cleanup of the various cases handled in the size calculation.

Resolves #18598

Authors:
  - Vyas Ramasubramani (https://github.com/vyasr)

Approvers:
  - Matthew Roeschke (https://github.com/mroeschke)
  - Matthew Murray (https://github.com/Matt711)

URL: #18734

Author: vyasr

python/pylibcudf/pylibcudf/column.pyx

@@ -86,25 +86,24 @@ cdef class OwnerWithCAI:
     cdef create(column_view cv, object owner):
         obj = OwnerWithCAI()
         obj.owner = owner
-        cdef int size
+        # The default size of 0 will be applied for any type that stores data in the
+        # children (such that the parent size is 0).
+        size = 0
         cdef column_view offsets_column
         cdef unique_ptr[scalar] last_offset
         if cv.type().id() == type_id.EMPTY:
             size = cv.size()
         elif is_fixed_width(cv.type()):
-            size = cv.size() * cpp_size_of(cv.type())
+            # Cast to Python integers before multiplyling to avoid overflow.
+            size = int(cv.size()) * int(cpp_size_of(cv.type()))
         elif cv.type().id() == type_id.STRING:
-            # The size of the character array in the parent is the offsets size
-            num_children = cv.num_children()
-            size = 0
-            # A strings column with no children is created for empty/all null
-            if num_children:
+            # A strings column with no children is created for empty/all null, in which
+            # case the size remains 0. Otherwise, the size of the character array stored
+            # in the parent is the last offset in the offsets child.
+            if cv.num_children():
                 offsets_column = cv.child(0)
                 last_offset = get_element(offsets_column, offsets_column.size() - 1)
                 size = (<numeric_scalar[size_type] *> last_offset.get()).value()
-        else:
-            # All other types store data in the children, so the parent size is 0
-            size = 0
 
         obj.cai = {
             "shape": (size,),