[VL] Test-only draft: GlutenDeltaVariantWriteSuite fails without the fix (validates #12444)

[felipepessoto]

Warning

Draft — do not merge. This PR intentionally contains only the test, without the fix, to
validate that the test fails on an unpatched main. The actual fix lives in
#12444.

What changes are proposed in this pull request?

This PR adds only GlutenDeltaVariantWriteSuite (backends-velox, delta-4.0) — the regression
test from #12444 — without the accompanying GlutenOptimisticTransaction guard. Its sole
purpose is to prove the test is an effective regression test: on an unpatched main, with the
native Delta write path enabled, writing a variant column throws
UnsupportedOperationException: Unsupported data type: variant, so the suite is expected to fail
on this branch.

The spark-test-spark40 job in velox_backend_x86.yml runs
-Pspark-4.0 -Pscala-2.13 -Pbackends-velox -Pdelta -Pspark-ut (no suite tags exclude it), so it
includes this suite. That job is expected to fail here and to pass once #12444's fix is applied.

How was this patch tested?

Not applicable — this branch is expected to fail CI by design. See #12444 for the fix together with
the passing tests.

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

Generated-by: GitHub Copilot CLI (Claude Opus 4.8)

[felipepessoto]

https://github.com/apache/gluten/actions/runs/28695499863/job/85112893664?pr=12445

[felipepessoto]

✅ CI validation: the tests fail without the fix, for the expected reason

This draft intentionally contains only the test suite, without the fix. The spark-test-spark40 and spark-test-spark41 jobs failed exactly as intended — confirming GlutenDeltaVariantWriteSuite is an effective regression guard.

All three tests fail on unpatched main, each with the identical root cause the fix targets:

org.apache.spark.SparkException: Job aborted due to stage failure:
  java.lang.UnsupportedOperationException: Unsupported data type: variant
    at org.apache.spark.sql.utils.SparkArrowUtil$.toArrowType(SparkArrowUtil.scala:60)
    at org.apache.spark.sql.utils.SparkArrowUtil$.toArrowField(SparkArrowUtil.scala:137)
    at org.apache.spark.sql.utils.SparkArrowUtil$.$anonfun$toArrowSchema$1(SparkArrowUtil.scala:167)
    ... (RowToVeloxColumnarExec — the offloaded native Delta write path)

Workflow	Job	Suite result
Velox Backend (x86)	spark-test-spark40	❌ 3/3 tests failed (variant)
Velox Backend (x86)	spark-test-spark41	❌ 3/3 tests failed (variant)
Velox Backend (Enhanced)	spark-test-spark40	❌ 3/3 tests failed (variant)