DDL grammar: writes_table on table-name slots for column narrowing

Handoff-12 §2.2: the DDL TABLE_NAME_EXISTING slot and the
relationship-endpoint table idents didn't set writes_table, so
column-name slots downstream (drop/rename/change column; relationship
qualified columns) couldn't narrow to the active table — candidates
leaked from every table. Set writes_table: true on TABLE_NAME_EXISTING
and on DR_PARENT/DR_CHILD/AR_PARENT/AR_CHILD table idents. The
deliberately-documenting completion test now asserts per-table
narrowing.

src/completion.rs

@@ -1177,8 +1177,11 @@ mod tests {
 
     #[test]
     fn drop_column_from_offers_only_current_table_columns() {
-        // The drop-column path also writes_table → narrowed
-        // columns should appear here too.
+        // The DDL `TABLE_NAME_EXISTING` slot now sets
+        // `writes_table` (handoff-13 §2.2 fix), so the
+        // column-name slot after the table name narrows to that
+        // table's columns. `OrderTotal` belongs to no table in
+        // this cache's `table_columns`, so it must not leak.
         use crate::dsl::types::Type;
         let mut cache = schema_with_table(
             "Customers",
@@ -1187,13 +1190,12 @@ mod tests {
         cache.columns.push("OrderTotal".to_string());
         let cs =
             cands_with("drop column from Customers: ", 28, &cache);
-        // Note: drop column's table-name slot doesn't set
-        // writes_table today (DDL paths don't carry Phase D
-        // table-column resolution yet). Falls back to global
-        // cache.columns, which is the documented schemaless
-        // fallback. Either narrowed-or-flat is acceptable; the
-        // test just confirms valid columns appear.
         assert!(cs.contains(&"Email".to_string()), "got {cs:?}");
+        assert!(cs.contains(&"id".to_string()), "got {cs:?}");
+        assert!(
+            !cs.contains(&"OrderTotal".to_string()),
+            "OrderTotal (not a Customers column) must not leak: got {cs:?}",
+        );
     }
 
     #[test]