Feature request: Support scaling across samples for univariate and multivariate time series

[AnnaBobasheva]

Is your feature request related to a problem? Please describe. Currently, tslearn scalers (e.g., TimeSeriesScalerMinMax) scale each sample time series individually. This removes differences in absolute values between samples, which can be problematic when those differences are meaningful.

Describe the solution you'd like It would be good to have an option in tslearn.preprocessing.TimeSeriesScalerMinMax and TimeSeriesScalerMeanVariance that scales each feature (variable) across the entire dataset, similar to how sklearn.MinMaxScaler works. This would preserve differences between samples for each feature, which is useful when absolute values matter.

Additionally, it would be helpful to support scaling across all features and all samples for multivariate time series, to allow for global scaling when appropriate (e.g., when all variables are on the same scale).

Ideally, the scaler should support both per-sample, per-feature, and per-dataset scaling modes.