RCR UB-Tree Multi-Version Management

The multi-version management of row consistency read (RCR) B-tree is based on data row levels. XIDs are recorded in data rows, which increases the key size, and the index expands by 5% to 20%. The latest and historical versions are on the B-tree, and the index does not record the undo information. Keys are inserted or deleted in the sequence of key + TID. Tuples with the same index column are sorted based on their TIDs as the second sorting keyword, and Xmin and Xmax are added to the end of the key. During index splitting, multi-version information is migrated with key migration.