Implement an efficient algorithm for finding differences between two versions (checkpoints) of the same btree. The high-level approach should be:

• If the address cookie of two subtrees are identical, skip them.
• If the address cookies are different, descend into them.
• At the leaf page level, compare each key/value pair individually.

This approach efficiently skips large parts of the btrees that are identical. Note that this requires:

• Precise checkpoints, as we'd like to avoid RTS and history store lookups.
• No prepared keys, as we'd like to avoid WT_PREPARE_CONFLICT.
• If necessary, also all pages being clean, so that the address cookie comparison works.

The motivating example for this work is WT-16188, so that we can efficiently determine which metadata entries changed in a shared metadata table between two checkpoints.