boxcox cannot de-exponentiate in a pipeline.

[adamd1985]

Describe the bug

[pmdarima\preprocessing\endog\boxcox.py:165](file:///C:/Users/worker/.conda/envs/tjoe/lib/site-packages/pmdarima/preprocessing/endog/boxcox.py:165): RuntimeWarning: invalid value encountered in power
  de_exp = numer ** (1. / lam1)  # de-exponentiate

To Reproduce

Happens all the time with pipelines, unless lambda1 is 0, this is an example setup:

from pmdarima.pipeline import Pipeline
from pmdarima.arima import auto_arima

model = Pipeline([
    ('boxcox', BoxCoxEndogTransformer(lmbda=0)),
    ('arima', auto_arima(y_train['Close'], d=n_diffs, trace=True, stepwise=True,
     seasonal=True, m=12, stationary=False, suppress_warnings=True, error_action="ignore", n_jobs=1))
])
model.fit(y_train['Close'])
...
c, conf_int = model.predict(
        n_periods=1, return_conf_int=True, inverse_transform=True)

Doesn't happen when called directly.

Use any large timeseries (e.g. a market index) to replicate, e.g.:

[{"FromDate":"2023-08-08T03:15:00Z","Open":15367.3,"High":15369.1,"Low":15349.2,"Close":15351.8,"Volume":null},{"FromDate":"2023-08-08T03:30:00Z","Open":15352.0,"High":15355.7,"Low":15348.6,"Close":15352.6,"Volume":null},{"FromDate":"2023-08-08T03:45:00Z","Open":15352.2,"High":15360.1,"Low":15351.7,"Close":15353.3,"Volume":null},{"FromDate":"2023-08-08T04:00:00Z","Open":15353.2,"High":15356.1,"Low":15350.8,"Close":15355.5,"Volume":null},{"FromDate":"2023-08-08T04:15:00Z","Open":15355.6,"High":15360.0,"Low":15354.2,"Close":15358.1,"Volume":null},
...
]

Versions

version: pmdarima  2.0.3  pypi_0 pypi

Expected Behavior

No errors, de-Exp happens correctly.

Actual Behavior

no de-exp happens, predicted value is correct but normalized.

Additional Context

No response