CNDB-15423: Make TopKProcessor lazily load queryVector

[michaeljmarshall]

What is the issue

Fixes: https://github.com/riptano/cndb/issues/15423
CNDB test: https://github.com/riptano/cndb/pull/15427

What does this PR fix and why was it fixed

This PR fixes a subtle bug exposed by the FeaturesVersionSupportCATest.testANN test, as well as other similarly named FeaturesVersionSupport*Test tests. The central issue is whether to create the queryVector object. The currently logic does so only when useSyntheticScore() is false. However, if a peer has it set to false and this node has it on true, we'll end up with an easily avoided NPE. I propose instead that we lazily load the vector when it is needed. This removes an unnecessary switch as well as removing an unnecessary ordering requirement on upgrades.

java.lang.NullPointerException
	at io.github.jbellis.jvector.vector.VectorUtil.squareL2Distance(VectorUtil.java:85)
	at io.github.jbellis.jvector.vector.VectorSimilarityFunction$1.compare(VectorSimilarityFunction.java:40)
	at org.apache.cassandra.index.sai.plan.TopKProcessor.getScoreForRow(TopKProcessor.java:333)
	at org.apache.cassandra.index.sai.plan.TopKProcessor.processScoredPartition(TopKProcessor.java:276)
	at org.apache.cassandra.index.sai.plan.TopKProcessor.reorder(TopKProcessor.java:145)
	at org.apache.cassandra.index.sai.plan.StorageAttachedIndexQueryPlan.lambda$postProcessor$2(StorageAttachedIndexQueryPlan.java:233)
	at org.apache.cassandra.db.ReadCommand.postReconciliationProcessing(ReadCommand.java:566)
	at org.apache.cassandra.service.reads.range.RangeCommands.partitions(RangeCommands.java:60)

[github-actions]

Checklist before you submit for review

• This PR adheres to the Definition of Done
• Make sure there is a PR in the CNDB project updating the Converged Cassandra version
• Use NoSpamLogger for log lines that may appear frequently in the logs
• Verify test results on Butler
• Test coverage for new/modified code is > 80%
• Proper code formatting
• Proper title for each commit staring with the project-issue number, like CNDB-1234
• Each commit has a meaningful description
• Each commit is not very long and contains related changes
• Renames, moves and reformatting are in distinct commits
• All new files should contain the DataStax copyright header instead of the Apache License one

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[cassci-bot]

✔️ Build ds-cassandra-pr-gate/PR-2007 approved by Butler

---

Approved by Butler
See build details here

[eolivelli]

Lgtm
Great catch.
I don't think we need additional unit tests in this case.
This change is actually fixing broken tests

[k-rus]

To me it looks good. My understanding is that the condition on the synthetic score was added by @adelapena, thus I mention him if he wants to see the fix in retrospective.