Extend the file: cursor's implementation of next_raw_key_value_batch so that it returns up to count key-value pairs from the same page. If the cursor is already positioned on a page, and there are more key-value pairs available, the function should return those. If the cursor is not positioned, it should behave in a way that is functionally equivalent to calling WT_CURSOR::next, following by the batch scan.
At this point, the cursor implementation does not need to consider prefix compression or WT_MODIFY; it should just fill in the provided key/value output arrays with WT_ITEMs that point to the btree page. We will handle the other cases in separate tickets.
As a part of this ticket, we should also include new functionality tests to ensure that the cursor is returning more than one key-value pair at a time whenever possible.