FIX: align DAE metadata with wide bindings

Use the Unicode-safe MAX boundary for VARCHAR data-at-execution metadata and compare the wide fetch regression by row values rather than generated Row identity.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

Author: gargsaumya

mssql_python/pybind/param_detect.hpp

@@ -144,7 +144,7 @@ inline SQLULEN DAEColumnSize(SQLSMALLINT sqlType, SQLULEN actualSize) {
     switch (sqlType) {
         case SQL_CHAR:
         case SQL_VARCHAR:
-            return actualSize > MAX_INLINE_BINARY ? 0 : actualSize;
+            return actualSize > MAX_INLINE_CHAR ? 0 : actualSize;
         case SQL_WCHAR:
         case SQL_WVARCHAR:
             return actualSize > MAX_INLINE_CHAR ? 0 : actualSize;

tests/test_004_cursor.py

@@ -1748,7 +1748,7 @@ def test_fetchmany_rejects_excessive_native_buffer(cursor):
 
 def test_fetchall_clamps_wide_result_batch_to_native_buffer_budget(cursor):
     cursor.execute("SELECT " + ", ".join("CAST(N'x' AS NVARCHAR(4000))" for _ in range(34)))
-    assert cursor.fetchall() == [("x",) * 34]
+    assert [tuple(row) for row in cursor.fetchall()] == [("x",) * 34]
 
 
 def test_description(cursor):