[P1] Add unit test for vanilla intervention at different positions and streams

[frankaging]

Description:
Adding tests when using vanilla intervention on the GPT-2 model at different streams. This PR focuses on position-based intervention. It should cover a single position as well as multiple positions. It should also cover different layers.

[frankaging]

Testing Done:

...
  File "/juice/scr/wuzhengx/align-transformers/models/mlp/modelings_mlp.py", line 60, in forward
    self.act(
  File "/u/nlp/anaconda/main/anaconda3/envs/wuzhengx-bootleg/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1545, in _call_impl
    hook_result = hook(self, args, kwargs, result)
  File "/juice/scr/wuzhengx/align-transformers/models/alignable_base.py", line 604, in hook_callback
    selected_output = self._gather_intervention_output(
  File "/juice/scr/wuzhengx/align-transformers/models/alignable_base.py", line 392, in _gather_intervention_output
    selected_output = gather_neurons(
  File "/juice/scr/wuzhengx/align-transformers/models/modeling_utils.py", line 284, in gather_neurons
    tensor_output = torch.gather(
 (Triggered internally at ../torch/csrc/autograd/python_anomaly_mode.cpp:114.)
  Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass
....
----------------------------------------------------------------------
Ran 13 tests in 4.751s

OK