fix: isinstance narrowing regression with dynamic tuple argument

When isinstance is called with a dynamically-computed tuple of types
(e.g. isinstance(exc, tuple(expected_excs))), the second argument's
stored type gets widened to match the _ClassInfo recursive type alias
(type | types.UnionType | tuple[_ClassInfo, ...]). The union handler
in get_type_range_of_type then decomposes this into individual members,
producing TypeRange(object, ...) from bare 'type', which incorrectly
narrows the expression to 'object' instead of keeping its existing type.

Fix by checking if the simplified union result is just 'object' — which
indicates we've lost type precision — and returning None to fall back to
keeping the current type, matching the v1.19 behavior.

Also handle None sub-items explicitly (propagate uncertainty) and filter
out UninhabitedType entries before simplifying.

Fixes #21181

Signed-off-by: bahtya <bahtyar153@qq.com>

Author: Bahtya

mypy/checker.py

@@ -8028,7 +8028,17 @@ def get_type_range_of_type(self, typ: Type) -> TypeRange | None:
 
         if isinstance(typ, UnionType):
             type_ranges = [self.get_type_range_of_type(item) for item in typ.items]
-            item = make_simplified_union([t.item for t in type_ranges if t is not None])
+            if any(t is None for t in type_ranges):
+                return None
+            valid_ranges = [t for t in type_ranges if not isinstance(get_proper_type(t.item), UninhabitedType)]
+            if not valid_ranges:
+                return TypeRange(UninhabitedType(), is_upper_bound=False)
+            # If the only meaningful type we can extract is "object", we've lost
+            # type precision (e.g. from a widened _ClassInfo alias). Return None
+            # to avoid narrowing to a useless type.
+            item = make_simplified_union([t.item for t in valid_ranges])
+            if isinstance(get_proper_type(item), Instance) and get_proper_type(item).type.fullname == "builtins.object":
+                return None
             return TypeRange(item, is_upper_bound=True)
         if isinstance(typ, FunctionLike) and typ.is_type_obj():
             # If a type is generic, `isinstance` can only narrow its variables to Any.