fix sql compatible null handling config work with runtime.properties

[clintropolis]

Description

This PR is an alternate to #8876, which was just awful, but I was afraid of introducing an if statement to NullHandling. replaceWithDefault in something that is called often on the query path. However, curiosity got the best of me, so I added a new benchmark schema to help determine how much of a performance hit introducing an if statement. It turns it that it isn't really noticeable, so I'm closing #8876 in favor of this PR instead.

before PR:

Benchmark                                              (defaultStrategy)  (initialBuckets)  (numProcessingThreads)  (numSegments)  (queryGranularity)  (rowsPerSegment)  (schemaAndQuery)  (vectorize)  Mode  Cnt      Score      Error  Units
GroupByBenchmark.queryMultiQueryableIndexWithSerde                    v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15   8610.792 ±  260.568  us/op
GroupByBenchmark.queryMultiQueryableIndexWithSpilling                 v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15   7769.964 ±  184.177  us/op
GroupByBenchmark.queryMultiQueryableIndexX                            v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15   8424.455 ±  268.964  us/op
GroupByBenchmark.querySingleIncrementalIndex                          v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15  15782.469 ± 3313.596  us/op
GroupByBenchmark.querySingleQueryableIndex                            v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15   3978.851 ±  284.614  us/op

after PR:

Benchmark                                              (defaultStrategy)  (initialBuckets)  (numProcessingThreads)  (numSegments)  (queryGranularity)  (rowsPerSegment)  (schemaAndQuery)  (vectorize)  Mode  Cnt      Score      Error  Units
GroupByBenchmark.queryMultiQueryableIndexWithSerde                    v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15   8104.835 ±  275.995  us/op
GroupByBenchmark.queryMultiQueryableIndexWithSpilling                 v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15   8215.000 ±  228.912  us/op
GroupByBenchmark.queryMultiQueryableIndexX                            v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15   8294.836 ±  189.671  us/op
GroupByBenchmark.querySingleIncrementalIndex                          v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15  16709.884 ± 3195.983  us/op
GroupByBenchmark.querySingleQueryableIndex                            v2                -1                       2              4                 all            100000    simple-nulls.A        false  avgt   15   3882.968 ±   93.530  us/op

---

This PR has:

• been self-reviewed.
• 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.
• added integration tests.
• been tested in a test Druid cluster.

---

Key changed/added classes in this PR

• NullHandling

[gianm]

I'm not sure having a volatile static is a good idea for something that will be used by multiple threads while processing every row. Your benchmark checked single-threaded performance but this could degrade under multithreading.

Could you please either look into this, or do a hybrid of your last two approaches (require explicit initialization in tests, of a non-volatile static, and if-guard it in methods that use it).

[clintropolis]

After discussion, maybe the explicit initialization is better in order to rid magic from our world, closing this PR and re-opening #8876