[Models] Add composable transformations to be applied before fitting a model

[AzulGarza]

Description

Transformations can often enhance the performance of certain types of models and data. We propose introducing a suite of composable transformations that can be applied to the time series before fitting the model. This might be implemented as follows:

from statsforecast.transformations import BoxCox, MinMax

sf = StatsForecast(
     models=[BoxCox(MinMax(AutoARIMA()))]
)

The new suite of transformations should be incorporated into the statsforecast/transformations.py module. The order of transformations should be from left to right; in the given example, BoxCox would be applied first, followed by MinMax, and finally the AutoARIMA model would be fitted on the fully transformed series.

Use case

No response