py/objstr: Make empty bytes object have a null-terminating byte.

Because a lot of string processing functions assume there is a null
terminating byte, so they can work in an efficient way.

Fixes issue #3334.

py/objstr.c

@@ -1973,8 +1973,8 @@ const mp_obj_type_t mp_type_bytes = {
     .locals_dict = (mp_obj_dict_t*)&str8_locals_dict,
 };
 
-// the zero-length bytes
-const mp_obj_str_t mp_const_empty_bytes_obj = {{&mp_type_bytes}, 0, 0, NULL};
+// The zero-length bytes object, with data that includes a null-terminating byte
+const mp_obj_str_t mp_const_empty_bytes_obj = {{&mp_type_bytes}, 0, 0, (const byte*)""};
 
 // Create a str/bytes object using the given data.  New memory is allocated and
 // the data is copied across.

tests/basics/bytes.py

@@ -8,6 +8,9 @@ print(b'\u1234')
 print(bytes())
 print(bytes(b'abc'))
 
+# make sure empty bytes is converted correctly
+print(str(bytes(), 'utf-8'))
+
 a = b"123"
 print(a)
 print(str(a))