Update analysis script to be standalone for reproducing the training dataset

[jcohenadad]

Currently, the processing script requires to be run on a dataset that was already processed by the scripts described in this PR (using this dataset as input).

We should, instead, move the scripts from the 'foreign' repos into this repos, to make a standalone package. See spine-generic/spine-generic#249 (comment)

[naga-karthik]

@sandrinebedard could you please take a look at this and close it when completed? Thanks!