[spark] Support AlwaysTrue and AlwaysFalse predicates in SparkV2FilterConverter

[kerwin-zk]

Purpose

Support AlwaysTrue and AlwaysFalse predicates in SparkV2FilterConverter

Tests

CI

[kerwin-zk]

cc @JingsongLi

[JingsongLi]

Why do we need to support this? Are there any beneficial queries?

[JingsongLi]

If we really need to introduce this, I prefer introducing TrueTransform and TrueFunction.

[kerwin-zk]

Why do we need to support this? Are there any beneficial queries?

@JingsongLi

DELETE FROM t1 WHERE pt IN (SELECT id FROM t2 WHERE n > 10) OR pt = 3
The subquery (n > 10) returns empty, evalSubquery replaces it with Literal(false). After translateFilterV2, this becomes: V2Or(ALWAYS_FALSE, V2EqualTo(pt, 3)). Without supporting ALWAYS_FALSE in SparkV2FilterConverter, the OR conversion would fail, falling back to the expensive row-level delete instead of partition dropping.