Fix initialization error in to_arrow for empty string views (#16033)

When converting an empty string view to arrow, we don't bother with copies from device, but rather create the arrow arrays directly. The offset buffer is therefore a singleton int32 array with zero in it.

Previously, the initialization of this array was incorrect, since mutable_data() returns a uint8_t pointer, and so setting the single element could leave 24 of the 32 bits uninitialized.

Fix this by using memset instead to zero out the full buffer.

Authors:
  - Lawrence Mitchell (https://github.com/wence-)

Approvers:
  - David Wendt (https://github.com/davidwendt)
  - Bradley Dice (https://github.com/bdice)

URL: #16033

cpp/src/interop/to_arrow.cu

@@ -292,9 +292,9 @@ std::shared_ptr<arrow::Array> dispatch_to_arrow::operator()<cudf::string_view>(
   auto child_arrays      = fetch_child_array(input_view, {{}, {}}, ar_mr, stream);
   if (child_arrays.empty()) {
     // Empty string will have only one value in offset of 4 bytes
-    auto tmp_offset_buffer               = allocate_arrow_buffer(4, ar_mr);
-    auto tmp_data_buffer                 = allocate_arrow_buffer(0, ar_mr);
-    tmp_offset_buffer->mutable_data()[0] = 0;
+    auto tmp_offset_buffer = allocate_arrow_buffer(sizeof(int32_t), ar_mr);
+    auto tmp_data_buffer   = allocate_arrow_buffer(0, ar_mr);
+    memset(tmp_offset_buffer->mutable_data(), 0, sizeof(int32_t));
 
     return std::make_shared<arrow::StringArray>(
       0, std::move(tmp_offset_buffer), std::move(tmp_data_buffer));