[FEATURE] Support batch friendly sort merge join

Is your feature request related to a problem?
The current logic of EnumerableMergeJoin in Calcite is retrieving all match rows for both side, and making Cartesian Product for them.

In some cases when the cardinality of the join key is very low, retrieving all rows with the same value on join key means fetching several batches from the index. However, since we always have query_size_limit in the end of PPL query or users may have their own limitation, and low cardinality means plenty of data explosion, the Cartesian Product of the first batch will usually satisfy the limit. Fetching batches more than requirements brings more query latency and memory usage.

What solution would you like?
Implements batch friendly sort merge join.

What alternatives have you considered?
A clear and concise description of any alternative solutions or features you've considered.

Do you have any additional context?
Add any other context or screenshots about the feature request here.

Start by tracing Calcite's EnumerableMergeJoin implementation and how OpenSearch SQL handles query_size_limit and index batches. Investigate the current retrieval of matching rows on both sides and its Cartesian-product behavior. Done means implementing a batch-friendly sort merge join that avoids unnecessary batch fetching while still satisfying the query limit.