[ENH] Implement CLeVer channel selection algorithm

[TonyBagnall]

Describe the feature or idea you want to propose

implement the CLeVer channel selection algorithm "CLeVer: A Feature Subset Selection Technique for Multivariate Time Series" https://ieeexplore.ieee.org/abstract/document/1490526

The Common principal component Loading based Variable subset selection method (CleVer) algorithm is PCA based approach that is adapted for channel selection for multivariate time series. CleVer uses PCA and clustering techniques to select dimensions. A PCA is performed independently on each instance, and the principal components are extracted. Next, all components that belong to the same class are combined to create common principal components through a process called Common Principal Component Analysis (cPCA). A proportion of the common components are used to create a feature space. These features are clustered, and the closest dimension to each centroid is chosen as selected dimensions.

Implement as a BaseCollectionTransformer in the channel_selection package