Fix aggregate queries without explicit grouping

[caglareker]

Fixes #3979

When continuing an aggregate query without an explicit GROUP BY clause, the code attempted to deserialize a group key even when no grouping key value existed, causing incorrect results.

Add null checks to handle the case where groupingKeyValue is null:

• Skip group key deserialization in the constructor if there's no explicit grouping
• Only include the group key in the serialized partial aggregation result when one exists

This ensures continued aggregate queries without grouping work correctly when resuming from a cursor.

[alecgrieser]

Thanks for the contribution! It seems like this could use a regression test covering the error case and then showing that it has been fixed by this change. I believe @hazefully had created a test case that showed this that used the yaml-tests framework.

The yaml-tests directory contains a bunch of queries, and it enables the queries to be executed both against the most recent server version but also against older versions in order to prevent against unexpected cross-version incompatibility problems.

The general process for creating a new file is to make a new file in yaml-tests/src/test/resources with your test case, and then to add a new method to YamlIntegrationTests that runs the tests for that file. You can look at some of the others (like, say, aggregate-index-tests.yamsql) to see how the file is formatted, but it allows the user to define a test schema and then queries that execute against it. I believe this bug only gets hit if an internal limit is hit, so IIRC, @hazefully had set the EXECUTION_SCANNED_ROWS_LIMIT option in the file in order to create a showcasing test case. (For an example of how we set connection options, you can look at, say, case-sensitivity.yamsql.)

There may be other ways to write a test for this. For example, FDBStreamAggregationTest has a bunch of other tests for the streaming aggregate code. It may be that we can add a test there and get more-or-less the same coverage

[caglareker]

added a yaml-tests case for ungrouped aggregate queries with continuations. uses maxRows=1 to force the continuation path that was hitting the NPE.