Correct why a script over a flat_object is refused

The comment claimed the field type does not support Painless for retrieving
subfield values. That is a limitation of doc-value access; this engine's
script reads _source, and it does retrieve flat_object leaves correctly --
a sort over one returns the right order. The reason to refuse is the cost
of that read, not an inability, and the types above are read from _source
on purpose, so the comment now says which of the two this is.

Signed-off-by: Jialiang Liang <jiallian@amazon.com>

Author: RyanL1997

opensearch/src/main/java/org/opensearch/sql/opensearch/storage/serde/RexStandardizer.java

@@ -117,9 +117,10 @@ public RexNode visitInputRef(RexInputRef inputRef, ScriptParameterHelper helper)
             ? null
             : OpenSearchTextType.toKeywordSubField(field.getName(), exprType);
     if (exprType instanceof OpenSearchFlatObjectType) {
-      // A flat_object has no usable doc values (each is the folded path=value term), so a script
-      // could only read it from _source, which means opening every record -- and the field type
-      // does not support Painless for retrieving subfield values either.
+      // Each leaf is indexed as one folded path=value term, so a flat_object has no doc values a
+      // script could read: it would have to open _source for every record it is asked about, the
+      // cost the field type exists to avoid. A script over one is refused rather than built, unlike
+      // the no-doc-values types above, which are read from _source that way on purpose.
       throw new UnsupportedScriptException(
           "A flat_object field cannot be read by a pushed-down script: " + field.getName());
     }