[MINOR][SQL] Centralize static and runtime filter eligibility checks

[szehon-ho]

What changes were proposed in this pull request?

This PR centralizes filter eligibility checks in PushDownUtils:

• Static filters are partitioned by determinism once and reused by the SupportsPushDownV2Filters and SupportsPushDownCatalystFilters paths. The shared value is lazy, so the legacy SupportsPushDownFilters path remains unchanged.
• Runtime filters are screened once before dispatching to the SupportsRuntimeV2Filtering or SupportsRuntimeCatalystFiltering path.

It also updates the runtime-filter method documentation to describe the enforced behavior instead of requiring callers to satisfy a precondition.

Why are the changes needed?

Filter eligibility is independent of the connector interface. Applying each shared check before interface dispatch removes duplicated screening and keeps equivalent connector paths aligned.

Does this PR introduce any user-facing change?

No.

How was this patch tested?

No new tests were added because existing tests cover non-deterministic static and runtime filters for the affected connector paths.

The following focused runtime-filter tests passed:

SPARK_LOCAL_IP=127.0.0.1 build/sbt 'sql/testOnly org.apache.spark.sql.connector.DataSourceV2SQLSuiteV2Filter -- -z "SPARK-58207: non-deterministic scalar subquery filters are not pushed into runtimeFilters"'

SPARK_LOCAL_IP=127.0.0.1 build/sbt 'sql/testOnly org.apache.spark.sql.connector.DataSourceV2CatalystRuntimeFilterSuite -- -z "non-deterministic predicate on fully pushed attributes"'

Also ran dev/lint-scala.

Was this patch authored or co-authored using generative AI tooling?

Generated-by: OpenAI Codex (GPT-5)