chore: add SQL tests for FIRST/LAST aggregates

[comphead]

Which issue does this PR close?

Closes #1630 #1646 .

Rationale for this change

Summary

Adds comprehensive SQL file tests for first and last aggregation functions with IGNORE NULLS semantics.

Motivation

The existing first_last.sql test file had only 3 basic queries covering first/last without any IGNORE NULLS coverage. The Scala test suite (CometAggregateSuite) has some IGNORE NULLS tests
but several are disabled (TODO #1646). SQL file tests are the preferred approach for expression testing in Comet and provide better coverage across the ConfigMatrix (dictionary encoding on/off).

What's added

Tests are organized into three sections: FIRST, LAST, and FIRST + LAST combined, each covering:

• Basic behavior — default (respect nulls) and IGNORE NULLS
• All-null groups — verifies NULL is returned when no non-null values exist
• Empty table — zero-row input
• Single row — one non-null value, and one null value
• Multiple data types — int, bigint, double, string, boolean (expect_fallback due to SortAggregate)
• Null positions — nulls at the beginning, middle, and end of groups
• Boolean parameter form — first(val, true) / first(val, false) equivalence with IGNORE NULLS syntax
• Expressions producing nulls — IF(val % 2 = 0, NULL, val) inside aggregation
• Mixed aggregations — first/last IGNORE NULLS alongside regular first/last, count, sum
• HAVING clause — filtering groups based on IGNORE NULLS result
• Decimal type — DECIMAL(10,2)
• Date type — DATE values
• Large group (1000 rows) — exercises multi-batch aggregation with sparse non-null values (every 10th row)

What changes are included in this PR?

How are these changes tested?

[kazuyukitanimura]

Thanks @comphead I think we cannot close #1630 yet because we are referring this issue e.g. docs/source/user-guide/latest/compatibility.md

The test itself looks good

[mbutrovich]

Thanks @comphead! Always good to have more tests.

[comphead]

Thanks @comphead I think we cannot close #1630 yet because we are referring this issue e.g. docs/source/user-guide/latest/compatibility.md

I think we can close #1630 as it referred to not using IGNORE NULLS flag, which is fixed now. However we need to track somewhere the deterministic feature, if we can actually reproduce it. WDYT?

[kazuyukitanimura]

Thanks @comphead I think we cannot close #1630 yet because we are referring this issue e.g. docs/source/user-guide/latest/compatibility.md

I think we can close #1630 as it referred to not using IGNORE NULLS flag, which is fixed now. However we need to track somewhere the deterministic feature, if we can actually reproduce it. WDYT?

Oh I see. You are right, we can close it. Let's just update the docs though
docs/source/user-guide/latest/compatibility.md
docs/source/user-guide/latest/expressions.md

[comphead]

I'll check also #1646 if everything works, we can just close 2 tickets and remove deterministic warning, otherwise we can refer to 1646

[kazuyukitanimura]

pending ci

[comphead]

Thanks everyone for review