[BUG] tslearn warpper bug when save_transformed_data is True

[TonyBagnall]

Describe the bug

In fit, when save_transformed_data=True, we save X_t into self.transformed_data, but X_t is just the tslearn-formatted input, not the shapelet transform.

Then get_transform(X) ignores its X argument and returns self.transformed_data_, and get_locations(X) ignores its X argument and calls self.clf_.locate(self.transformed_data_). So we are not actually saving the transform, and we are not actually using the X passed to those methods.

this is an addition to #3379 where it crashes with version 0.8.0

Steps/Code to reproduce the bug

import numpy as np

from aeon.classification.shapelet_based import LearningShapeletClassifier
from aeon.testing.data_generation import make_example_3d_numpy

X_train = make_example_3d_numpy(
    n_cases=10, n_channels=1, n_timepoints=20, random_state=0, return_y=False
)
X_test = make_example_3d_numpy(
    n_cases=3, n_channels=1, n_timepoints=20, random_state=1, return_y=False
)
y_train = np.array([0, 1] * 5)

clf = LearningShapeletClassifier(
    max_iter=10, total_lengths=1, save_transformed_data=True, random_state=0
)
clf.fit(X_train, y_train)

# Bug 1: get_transform(X) ignores X and returns cached training data
Xt_test = clf.get_transform(X_test)
assert np.array_equal(Xt_test, clf.transformed_data_)

# Bug 2: returned object has training-set case count, not test-set case count
assert Xt_test.shape[0] == X_train.shape[0]
assert Xt_test.shape[0] != X_test.shape[0]

# Bug 3: get_locations(X) ignores X as well
loc_train = clf.get_locations(X_train)
loc_test = clf.get_locations(X_test)
assert np.array_equal(loc_train, loc_test)

Expected results

save transform not X

Actual results

saves X not transform

Versions

No response