BUG : custom features + Sentinel 1

If Sentinel-1 is enable and getter methods, such as get_interpolated_Sentinel1_ASC_vh, are used then the following error occurs at iota2's start :

AttributeError: 'custom_numpy_features' object has no attribute 'get_interpolated_Sentinel1_ASC_vh'

triggered by the function test_user_feature_with_fake_data().

This function prepare fake data to feed a custom_numpy_features object and check if the external features is usable.

However most of the custom_numpy_features class design is based on available dates for every enabled sensors. At the iota's beginning, there is no Sentinel-1 tiled data and therefore no available dates leading to the observed error. Consequently, we must preprocess all of Sentinel-1 data before checking if the user function is usable.

A solution can be the use of the prelaunch_function functionality as i2_classification.py#L222 to check the external features consistency between two step containers.

To perform this check asap, I recommend to encapsulate every sensors preprocess in a dedicated containers -> Minor impact on parallelization granularity : every sensors preprocessing must be achieve before doing anything else.