[BUG] `LeftSTAMPi` output has incorrect shape

[MatthewMiddlehurst]

Describe the bug

Length of the output should be the same as n_timepoints, seems to be smaller in reality. This causes check_series_anomaly_detector_output to fail. Currently skipped with "LeftSTAMPi": ["check_series_anomaly_detector_output"] in testing_config.

Steps/Code to reproduce the bug

CI output:

estimator = LeftSTAMPi(), datatype = 'UnivariateSeries-None'

    def check_series_anomaly_detector_output(estimator, datatype):
        """Test the series anomaly detector output on valid data."""
        estimator = _clone_estimator(estimator)
    
        estimator.fit(
            FULL_TEST_DATA_DICT[datatype]["train"][0],
            FULL_TEST_DATA_DICT[datatype]["train"][1],
        )
    
        y_pred = estimator.predict(FULL_TEST_DATA_DICT[datatype]["test"][0])
        assert isinstance(y_pred, np.ndarray)
        # series need n_timepoints predictions
>       assert len(y_pred) == FULL_TEST_DATA_DICT[datatype]["test"][0].shape[1]
E       AssertionError

datatype   = 'UnivariateSeries-None'
estimator  = LeftSTAMPi()
y_pred     = array([0.        , 0.        , 0.        , 0.03453322, 0.07458139,
       0.40225739, 0.78719632, 0.95601876, 0.865249...93, 0.22668   , 0.19543261, 0.08394415, 0.05453156,
       0.07175253, 0.04052627, 0.03479472, 0.02411278, 0.04349395])

Expected results

Test does not fail.

Actual results

Test fails.

Versions

No response