Enable identifier expansion to ensure that casted args in agg functions are passed on

Signed-off-by: Yuanchun Shen <yuanchu@amazon.com>

Author: yuancu

core/src/main/java/org/opensearch/sql/calcite/CalcitePlanContext.java

@@ -145,7 +145,12 @@ public SqlValidator getValidator() {
               // Use lenient conformance for PPL compatibility
               .withConformance(OpenSearchSparkSqlDialect.DEFAULT.getConformance())
               // Use Spark SQL's NULL collation (NULLs sorted LOW/FIRST)
-              .withDefaultNullCollation(NullCollation.LOW);
+              .withDefaultNullCollation(NullCollation.LOW)
+              // This ensures that coerced arguments are replaced with cast version in sql select
+              // list because coercion is performed during select list expansion during sql
+              // validation. Affects 4356.yml
+              // See SqlValidatorImpl#validateSelectList and AggConverter#translateAgg
+              .withIdentifierExpansion(true);
       validator =
           PplValidator.create(
               statement,

core/src/main/java/org/opensearch/sql/executor/QueryService.java

@@ -318,7 +318,7 @@ private RelNode validate(RelNode relNode, CalcitePlanContext context) {
     SqlNode rewritten = root.accept(new SqlRewriteShuttle());
     SqlValidator validator = context.getValidator();
     try {
-      validator.validate(Objects.requireNonNull(rewritten));
+        validator.validate(Objects.requireNonNull(rewritten));
     } catch (CalciteContextException e) {
       if (ValidationUtils.tolerantValidationException(e)) {
         return relNode;