Expand testEnumUnionString comment with upstream-bug context

[YogeshKothari26]

What changes are proposed in this pull request, and why are they necessary?

Documents context for the disabled testEnumUnionString test (disabled in #585):

• The failure is an upstream Hive 2.3.9 parser bug — SemanticAnalyzer throws AssertionError in UnparseTranslator.addTranslation during CREATE VIEW parse for a UNION ALL between an Avro enum column and a string column.
• Coral's mergeUnionSchema (added in #282) is unaffected and still produces a correct STRING-typed schema for already-created views at translation time.
• The bug fires at CREATE VIEW parse time on Hive 2.3.9. The case stays unsupported in Coral CI/i-tests (which embed Hive 2.3.9).
• Test will be re-enabled once the upstream Hive parser issue is resolved.

Comment-only change.

How was this patch tested?

./gradlew :coral-schema:spotlessCheck — PASS. No code change.

[aastha25]

if its a parser bug, does is mean its a hive 2.3 client side bug? we have internal view creation pipelines which use spark (and spark is internally using hive 2.3 client libs). is that job impacted or not?

[YogeshKothari26]

@aastha25 — verified Spark CREATE VIEW path is not impacted by this bug.

Test: Created a Hive-backed Avro table with an enum column, ran the equivalent CREATE VIEW v AS SELECT enum_field FROM t UNION ALL SELECT 'literal_string' FROM t on Spark 3.5 with spark.sql.hive.metastore.version = 2.3.9.1. View created successfully, no AssertionError.

Why Spark is safe: Spark parses CREATE VIEW DDL with its own SqlBaseParser and stores the view body verbatim via HiveExternalCatalog.createTable → Hive.createTable (metadata API only). Hive's SemanticAnalyzer / UnparseTranslator — where this bug lives — are never invoked in Spark's flow. DESCRIBE EXTENDED confirms this empirically: View Text and View Original Text are byte-identical, meaning Spark stored the SQL as-is without re-canonicalizing. If UnparseTranslator had run, View Text would be the re-serialized form.

The bug fires only when Hive's parser is directly invoked (e.g. Hive CLI, embedded Hive in Coral's HiveTester). Spark-based view creation pipelines remain unaffected.