Add benchmark suite for MSQ window functions

[Akshat-Jain]

Description

This PR adds a benchmark suite for MSQ window function queries.

A sample run on my local gave the following results:

Benchmark                                                         (maxNumTasks)  (rowsPerSegment)  Mode  Cnt      Score      Error  Units
MSQWindowFunctionsBenchmark.windowWithoutGroupBy                              2          20000000  avgt    5  96681.604 ± 2425.579  ms/op
MSQWindowFunctionsBenchmark.windowWithoutGroupBy                              5          20000000  avgt    5  94676.305 ± 4012.108  ms/op
MSQWindowFunctionsBenchmark.windowWithoutSorting                              2          20000000  avgt    5  17515.970 ±  498.066  ms/op
MSQWindowFunctionsBenchmark.windowWithoutSorting                              5          20000000  avgt    5  15996.262 ± 1552.218  ms/op
MSQWindowFunctionsBenchmark.multipleWindows                                   2          20000000  avgt    5   63215.499 ± 4722.604  ms/op
MSQWindowFunctionsBenchmark.multipleWindows                                   5          20000000  avgt    5   69287.847 ± 4985.326  ms/op
MSQWindowFunctionsBenchmark.windowWithHighCardinalityPartitionBy              2          20000000  avgt    5   69469.122 ± 3016.019  ms/op
MSQWindowFunctionsBenchmark.windowWithHighCardinalityPartitionBy              5          20000000  avgt    5   70951.896 ± 4343.354  ms/op
MSQWindowFunctionsBenchmark.windowWithLowCardinalityPartitionBy               2          20000000  avgt    5     507.584 ±  600.999  ms/op
MSQWindowFunctionsBenchmark.windowWithLowCardinalityPartitionBy               5          20000000  avgt    5     413.795 ±   38.195  ms/op
MSQWindowFunctionsBenchmark.windowWithSorting                                 2          20000000  avgt    5   16682.792 ±  239.561  ms/op
MSQWindowFunctionsBenchmark.windowWithSorting                                 5          20000000  avgt    5   16422.890 ±  225.643  ms/op

(Note: The above run was done with the changes of #17373, as otherwise I was running into the Channel has no capacity issue)

---

This PR has:

• been self-reviewed.
• added documentation for new or modified features or behaviors.
• a release note entry in the PR description.
• added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links.
• added or updated version, license, or notice information in licenses.yaml
• added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader.
• added unit tests or modified existing tests to cover new code paths, ensuring the threshold for code coverage is met.
• added integration tests.
• been tested in a test Druid cluster.

[adarshsanjeev]

Overall, this looks good to me. Few comments.

[adarshsanjeev]

I'm not very familiar with this, why is the change required to run all runsteps for this benchmark (and not others)?

[Akshat-Jain]

queryTestBuilder.results() didn't support MSQ apparently, since it only had the execute step, and not the ExtractResultsFactory step.

So, without this change, instead of getting the query results, we were getting the query ID. The ExtractResultsFactory step (added as a custom runner when declaring QueryTestBuilder) takes the query ID, and contacts the overlord client, and fetches the actual results.

(and not others)?

In the regular MSQ tests, we do testBuilder().run() which handled both steps already.