fix(admin): guard find_changed_cell against cross-table schema mismatch

When switching tables, the table-switch branch writes fresh data for the
new table, which re-fires the same callback with data_previous still
holding the old table's rows. find_changed_cell only checked row count
before diffing, so if the two tables happened to have the same number of
rows it would zip mismatched-schema dicts and report a spurious changed
cell (usually the PK column), producing a confusing red alert right after
switching tables.

src/admin/tables.py

@@ -144,6 +144,10 @@ def find_changed_cell(
 ) -> tuple[int, str, object, object] | None:
     if data_previous is None or len(data) != len(data_previous):
         return None
+    if not data or not data_previous:
+        return None
+    if set(data[0].keys()) != set(data_previous[0].keys()):
+        return None
     for i, (new_row, old_row) in enumerate(zip(data, data_previous)):
         for col, new_val in new_row.items():
             old_val = old_row.get(col)

tests/admin/test_tables.py

@@ -99,6 +99,16 @@ def test_find_changed_cell_returns_none_when_identical():
     assert tables.find_changed_cell(data, data_previous) is None
 
 
+def test_find_changed_cell_returns_none_when_schemas_differ():
+    # Mimics switching from "users" to "admin_actions" mid-callback: the two
+    # tables' row lists happen to have the same length but different
+    # columns, so this must not be mistaken for a genuine cell edit.
+    data = [{"id": 1, "email": "a@ex.fr"}]
+    data_previous = [{"id": 1, "admin_email": "x@ex.fr"}]
+
+    assert tables.find_changed_cell(data, data_previous) is None
+
+
 def test_find_changed_cell_returns_none_when_previous_is_none():
     assert tables.find_changed_cell([{"id": 1}], None) is None