KTR predict method doesn't consider single knot case.

[Eve-Chen]

Describe the bug
In the case when user want to fit a flat level (and build customized trend as regressors) by specifying level_segments=0, the KTR model is able to fit a single knot for the level component. However, the predict function produces IndexError because the underlying code assumes KTRModel.knots_tp_levels to have at least 2 components.

To Reproduce

import orbit
from orbit.models import KTR
from orbit.diagnostics.plot import plot_predicted_data, plot_predicted_components
from orbit.utils.dataset import load_electricity_demand

df = load_electricity_demand()
date_col = 'date'
response_col = 'electricity'
df[response_col] = np.log(df[response_col])
test_size=365
train_df=df[:-test_size]
test_df=df[-test_size:]
ktr = KTR(
    response_col=response_col,
    date_col=date_col,
    seed=2021,
    estimator='pyro-svi',
    level_segments=0,
    # bootstrap sampling to capture uncertainties
    n_bootstrap_draws=1e4,
    # pyro training config
    num_steps=301,
    message=100,
)
ktr.fit(train_df)
ktr.predict(test_df)

The code above gives IndexError: index -2 is out of bounds for axis 0 with size 1.

Expected behavior
Separate code paths for single and multiple level knots.

Screenshots
If applicable, add screenshots to help explain your problem.

Environment (please complete the following information):

• OS: Linux
• Python Version: 3.10.8
• Versions of Major Dependencies (pandas, scikit-learn, cython): pandas==1.5.3, scikit-learn==1.2.1

Additional context
Could we separate level and trend? I.e. Level is the dynamic intercept part, and trend is the dynamic slope part, which can extend into the future smoothly.