scd2_upsert now replays duplicate keys within a single batch in chronological order so every
change is stored as its own historical row.
The SCD demo notebook includes a "multiple updates per batch" walk-through and schema-evolution
examples for both SCD1 and SCD2 to visualize the resulting Delta output.
Optional schema evolution (allow_schema_evolution=True) lets Delta automatically add columns when
the upstream dataset introduces new attributes across both SCD helpers.
No action is required for existing pipelines. The helper automatically sequences inputs that have
not been pre-deduplicated and will continue to behave as before when one row per key is provided.
scd2_upsert partitions each batch by business key and load order so earlier versions are closed
before inserting later ones, retaining complete history at write time.
Noted the new intra-batch sequencing behavior in the SCD guide and notebook, including a dedicated
example that shows the resulting timeline of versions.