Tagging python transform class as experimental breaks tests

`As part of BEAM-7715 I tried to add the `@experimental()` annotation in the external.py `External` and generate_sequence transform too. However it broke the tests apparently on a typing issue related to the annotation.`

`Log:`
*15:53:42* File "/home/jenkins/jenkins-slave/workspace/beam_PreCommit_Python_PVR_Flink_Commit/src/sdks/python/apache_beam/runners/portability/flink_runner_test.py", line 172, in test_external_transforms*15:53:42* expansion_service=get_expansion_service()))}}
`*15:53:42* File "apache_beam/utils/annotations.py", line 146, in inner*15:53:42* return fnc(*args, **kwargs)**`
`*15:53:42* File "apache_beam/io/external/generate_sequence.py", line 56, in __init_`
`___*15:53:42* super(GenerateSequence, self).__init__()*15:53:42* TypeError: super() argument 1 must be type, not functionRef.`

Ref.
[https://builds.apache.org/job/beam_PreCommit_Python_PVR_Flink_Commit/3794/console](https://builds.apache.org/job/beam_PreCommit_Python_PVR_Flink_Commit/3794/console)

Imported from Jira [BEAM-7757](https://issues.apache.org/jira/browse/BEAM-7757). Original Jira may contain additional context.
Reported by: iemejia.