Support SQL aggregate FILTER (WHERE ...) clause in native execution

[yew1eb]

Problem

Queries using the SQL standard FILTER (WHERE ...) clause on aggregate functions currently fall back to non-native Spark execution. For example:

SELECT SUM(i) FILTER (WHERE flag = true) FROM t

This prevents users from getting the performance benefits of native execution for a commonly used SQL pattern.

Expected Behavior

FILTER (WHERE ...) on aggregates should be natively executed without falling back to Spark.

Current Behavior

The native converter throws an assertion error and falls back to HashAggregateExec.

Use Cases

• Conditional aggregation in a single pass (e.g., sum of revenue where status = 'completed')
• Multiple aggregates with different predicates in the same SELECT
• Standard SQL compatibility

Example Query

SELECT
  SUM(i) FILTER (WHERE flag = true)  AS sum_true,
  SUM(i) FILTER (WHERE flag = false) AS sum_false,
  AVG(d) FILTER (WHERE j > 20)       AS avg_d,
  COUNT(*) FILTER (WHERE i IS NULL)  AS cnt_null
FROM test_agg_filter