Some optimizations for constant blocks

[dnhatn]

This change introduces several optimizations for constant blocks:

1. When reading keyword fields that are index-sorted and the query range is large enough, we can return a constant block of BytesRef values to enable downstream optimizations.

2. Enable shortcuts for BytesRefBlockHash and BytesRefLongBlockHash when handling constant blocks. These optimizations are quick wins for time-series aggregations.

3. Enable shortcuts for CountAggregator and ValuesAggregator for constant blocks.

[dnhatn]

Here is the benchmark with this change:

Before:

{
    "operator": "TimeSeriesAggregationOperator[blockHash=BytesRefLongBlockHash{keys=[BytesRefKey[channel=3], LongKey[channel=2]], entries=546, size=56368b}, aggregators=[GroupingAggregator[aggregatorFunction=SumDoubleGroupingAggregatorFunction[channels=[4]], mode=INITIAL], GroupingAggregator[aggregatorFunction=CountGroupingAggregatorFunction[channels=[4]], mode=INITIAL], GroupingAggregator[aggregatorFunction=ValuesBytesRefGroupingAggregatorFunction[channels=[5]], mode=INITIAL]]]",
    "status": {
        "hash_nanos": 15198487, <-15ms
        "aggregation_nanos": 29086469, <- 29ms
        "pages_processed": 546,
        "rows_received": 982982,
        "rows_emitted": 546,
        "emit_nanos": 148580
    }
}

After:

{
    "operator": "TimeSeriesAggregationOperator[blockHash=BytesRefLongBlockHash{keys=[BytesRefKey[channel=3], LongKey[channel=2]], entries=546, size=56368b}, aggregators=[GroupingAggregator[aggregatorFunction=SumDoubleGroupingAggregatorFunction[channels=[4]], mode=INITIAL], GroupingAggregator[aggregatorFunction=CountGroupingAggregatorFunction[channels=[4]], mode=INITIAL], GroupingAggregator[aggregatorFunction=ValuesBytesRefGroupingAggregatorFunction[channels=[5]], mode=INITIAL]]]",
    "status": {
        "hash_nanos": 3291481, <- 3ms
        "aggregation_nanos": 22296535, <- 22ms
        "pages_processed": 546,
        "rows_received": 982982,
        "rows_emitted": 546,
        "emit_nanos": 130129
    }
}

[dnhatn]

Here, we choose to return an ordinal constant block instead of a direct constant block to ensure that ordinal optimizations are applied if the constant optimization is not available.

[elasticsearchmachine]

Hi @dnhatn, I've created a changelog YAML for you.

[martijnvg]

This speedup is amazing!

[martijnvg]

This looks good! I left two comments.

[elasticsearchmachine]

Pinging @elastic/es-analytical-engine (Team:Analytics)

[martijnvg]

LGTM 👍

[nik9000]

Should this be a ConstantBytesRefBlock?

[dnhatn]

Yeah, it should be but here we need to return an ordinal constant block instead of a direct constant block to ensure that ordinal optimizations are applied if the constant optimization is not available: #132456 (comment)

[nik9000]

Got it. It's worth adding a comment about why that's not a ConstantBlock I think.

[dnhatn]

It's worth adding a comment about why that's not a ConstantBlock I think.

++, I added a comment in 276759a.

[dnhatn]

@martijnvg @nik9000 Thank you!