If the values in the btree use WT_MODIFY, calling next_raw_key_value_batch should reconstruct the appropriate values in a separate buffer owned by the cursor. This should be analogous to how the function decodes keys encoded using prefix compression.
As a part of this ticket, we should create the corresponding functionality tests that use WT_CURSOR::modify when updating a table, and then ensuring that next_raw_key_value_batch returns the expected values.