Start working on distributed framework

[furlat]

All the functional methods or the at least their higher level wrappers should have a way to be executed in a distributed fashion using either:

1. https://modal.com/docs/guide/scale
2. Slurm in preparation of Leonardo
3. Something simpler to handle multiple gpus in the same lan/vpn

The first focus is being able to execute the stratified cv in parallel on cpu machines, next the embeddings using transformer models.