Issues with KNeighborsTimeSeriesClassifier with multivariate data

[TonyBagnall]

1. it is very slow (slower than HC2) and does not complete.
2. It is meant to work with MSM, but MSM is not capable of multivariate
3. Crashes for train preds on some problems. This may be a problem with generating train files not the classifier, but it does not happen with other algorithms (check!). Seems

File "/gpfs/home/ajb/Code/estimator-evaluation/sktime_estimator_evaluation/experiments/classification_experiments.py", line 129, in
predefined_resample=predefined_resample,
File "/gpfs/home/ajb/Code/sktime/sktime/benchmarking/experiments.py", line 536, in load_and_run_classification_experiment
test_file=build_test,
File "/gpfs/home/ajb/Code/sktime/sktime/benchmarking/experiments.py", line 392, in run_classification_experiment
train_preds = classifier.classes_[np.argmax(train_probs, axis=1)]
TypeError: only integer scalar arrays can be converted to a scalar index

this is concerned with generating train estimates with 10 fold CV, something going wrong there I think.